10 Breakthrough Technologies 2025 – Highlights from SXSW
Niall Firth from MIT presented the 10 Breakthrough Technologies 2025 at this year’s SXSW 2025 conference. Discover the ten innovations poised to reshape science, industry, and society. This presentation is also available as a podcast—tune in to explore the future today.
A showcase of ten breakthrough innovations poised to reshape science, industry, and society. From a next-generation observatory hunting dark matter to AI
systems transforming everything from web search to healthcare, these technologies were presented as the year’s top breakthroughs. Below is a comprehensive report in a journalistic style, with clear explanations, current developments, and potential challenges for each technology.
Table of Contents:
- Vera Rubin Observatory and Dark Matter Research
- Generative AI in Search Engines
- Small Language Models (SLMs)
- Cattle “Burping” Remedies (Methane Reduction)
- Robo-Taxis
- Cleaner Jet Fuel (Sustainable Aviation Fuel)
- Versatile AI-Powered Robots
- Long-Acting HIV Prevention Medication
- Green Steel
- Effective Stem Cell Therapies
You are currently seeing a placeholder content of Podigee. To access the actual content, click on the button below. Please note that data will be passed on to third-party providers.
More information1. Vera Rubin Observatory and Dark Matter Research
Figure: The Vera C. Rubin Observatory under construction on Cerro Pachón in Chile (2021).
Equipped with the world’s largest digital camera (3.2 gigapixels), this facility will survey the southern sky nightly in the Legacy Survey of Space and Time, aiming to map the universe in unprecedented detail .
Overview: The Vera C. Rubin Observatory is a cutting-edge astronomical observatory in northern Chile, expected to see “first light” in 2025 . Named after astronomer Vera Rubin – who provided the first compelling evidence of dark matter – it features an 8.4-meter telescope and the largest digital camera ever built for astronomy . Over a ten-year survey, the Rubin Observatory will repeatedly photograph the entire southern sky, creating a high-resolution time-lapse “movie” of the universe . Scientists hope this vast dataset will help answer fundamental questions about dark matter and dark energy, the mysterious components that make up most of the cosmos .
How It Works and Why It Matters:
The observatory’s extraordinary camera (3200 megapixels) and novel three-mirror telescope design enable rapid, wide-field imaging of faint objects . Every few nights it will cover the sky, amassing about 20 terabytes of images each night and an estimated 5,000 terabytes per year . This deluge of data will capture transient events (like supernovae or asteroid fly-bys) and reveal subtle patterns in galaxy motions. Crucially, the Rubin Observatory’s data will aid the quest to understand dark energy and dark matter – the invisible forces believed responsible for the universe’s accelerated expansion and for 85% of its matter . By mapping billions of galaxies and their distortions, researchers can infer the distribution of dark matter and test cosmological theories. If successful, this project could solve long-standing cosmic mysteries, marking “a new era of astronomy” and potentially a scientific revolution .
Recent Developments:
After decades of planning, construction is nearly complete. In April 2024, engineers finished building the giant camera at SLAC National Accelerator Laboratory . The camera can capture an area of sky 40 times the size of the full moon in a single exposure, with such high resolution that a single image would fill hundreds of HDTV screens . Testing has been successful – the observatory’s systems passed full integration tests, and “first light” (the first sky images) is anticipated in July 2025.
Already, science teams worldwide are preparing to mine the data for discoveries, from finding hazardous near-Earth asteroids to detecting the subtle gravitational lensing signals of dark matter. The excitement in the astronomy community is palpable, as Rubin promises to “bring the night sky to life” and yield a treasure trove of discoveries .
Challenges and Concerns:
Managing the enormous data stream is a major challenge – petabyte-scale databases and advanced algorithms are needed to store, process, and analyze the images in near-real-time. There are also concerns about satellite megaconstellations (like SpaceX’s Starlink) streaking through images and potentially compromising observations of faint objects. Additionally, while the observatory is expected to shed light on dark matter, there is no guarantee it will “solve” the dark matter puzzle; unexpected results could raise new questions.
Finally, as with any big science project, continued funding and smooth international collaboration are vital. Nonetheless, optimism is high that Vera Rubin Observatory will significantly advance our understanding of the universe’s dark components, fulfilling its promise as one of the most ambitious astronomy projects in history .
2. Generative AI in Search Engines
Overview: Traditional web search is undergoing a radical transformation with the integration of generative AI.
Tech giants like Google and Microsoft have rolled out AI-generated summary answers at the top of search results, fundamentally changing how we find information .
Instead of a list of links that require clicking, users can now get an instant answer or overview written by an AI. For example, Google’s Search Generative Experience and Microsoft’s AI-powered Bing use large language models (similar to ChatGPT) to provide a concise answer with cited sources.
This “zero-click” search paradigm means the search engine itself can directly answer many queries .
It’s a paradigm shift that could disrupt online business models – from how news and content sites get traffic to how search engines earn
ad revenue .
How It Works and Why It Matters:
These AI search systems work by drawing on vast web indexes and language model training to compose natural-language responses to user queries. When you ask a question, the AI summarizes relevant information from multiple sources into a single answer. The appeal is obvious: it’s fast and convenient – users “get relevant answers faster without having to click multiple results” . This matters because it could make searching more efficient and accessible. For users, AI-driven search offers the convenience of direct answers and the ability to ask follow-up questions in a conversational mode. It’s like having a knowledgeable assistant rather than just a directory of links.
From an industry perspective, however, this shift has enormous implications. AI summaries threaten to “take a bite out of click-through traffic” that websites rely on . A recent study found about 80% of consumers now use “zero-click” AI results in at least 40% of their searches, reducing organic web traffic by an estimated 15–25% . If users no longer click news articles or visit recipe blogs because the AI already gave them the answer, entire sectors – journalism, SEO, online advertising – may need to reinvent themselves . This is why some commentators are calling it “the end of the classic Internet”.
On the flip side, it pushes website owners to focus on higher-quality, in-depth content (since basic queries might be answered by AI) and new ways to attract readers.
Recent Developments:
In the past year, we’ve seen rapid deployment of these technologies. Microsoft’s Bing integrated OpenAI’s GPT-4 into search in early 2023, and by 2024 it reported millions of users engaging with the AI chat mode. Google followed by introducing AI “Search Overviews” for many queries, available to users in the U.S. as an experimental feature .
These overviews synthesize information and even include citations and links to the sources. The rollout hasn’t been entirely smooth – users and experts quickly noticed AI errors (“hallucinations”) in some answers, where the chatbot would state incorrect facts with confidence. For instance, Google’s Bard made a factual mistake in a demo that wiped out $100 billion in Alphabet’s market value, underscoring the stakes. Despite such hiccups, the integration continues. By late 2024, Google said about 60% of searches now end without a click to another site, due in part to AI summaries . Marketers are adapting: companies are exploring how to get their content featured in AI answers, and search engines are tweaking how they attribute and link to sources to keep the web ecosystem intact. It’s a fast-moving
space, with new AI search features and improvements (e.g. to factual accuracy) rolling out almost monthly.
Challenges and Concerns:
The rise of AI-driven search brings significant concerns. Misinformation and accuracy are top of the list – these language models can sometimes generate plausible-sounding but false information (known as hallucinations) . Unlike a traditional search result that simply shows existing content (which a user might cross-check), an AI might confidently present an answer that is partially or entirely incorrect, potentially misleading users. As one expert noted, “LLMs can confidently produce false statements…combining unrelated concepts in nonsensical ways” . This puts pressure on search companies to develop verification and citation methods and to clearly indicate uncertainty. There are also ethical and economic concerns: websites fear losing traffic and revenue, content creators worry about being scraped without credit, and there’s a broader worry that a few AI systems will intermediate most information flow (raising questions about bias and control).
Privacy is another aspect – these AI tools require enormous amounts of data and could inadvertently expose personal or sensitive information queried by users.
Lastly, the user trust issue looms: will people trust AI answers? Surveys show mixed reactions – while many enjoy the convenience, others remain skeptical and prefer to double-check facts independently. Regulators are beginning to pay attention, and the coming years will likely see debates on how to ensure AI search is transparent, accountable, and benefits the broader internet ecosystem, not just the companies deploying it.
3. Small Language Models (SLMs)
Overview: In contrast to the tech industry’s race toward ever-larger AI models, a counter-trend is emerging: Small Language Models (SLMs). These are more compact AI models designed to perform language and other AI tasks with far fewer parameters than the likes of GPT-4. While giants like OpenAI, Google, and Anthropic have been building models with hundreds of billions of parameters, SLMs focus on being specialized, efficient, and accessible . The idea is that bigger isn’t always better – for many applications, a smaller, fine-tuned model can achieve similar results with a fraction of the computational resources.
This trend represents a “return to efficiency” in AI development , and it’s gaining momentum as businesses seek AI solutions that they can deploy on their own hardware or devices, without relying on Big Tech’s cloud supercomputers.
How It Works and Why It Matters:
Small language models are essentially streamlined versions of the large language models (LLMs) that have grabbed headlines. They use architectures optimized to require less data and training time, often focusing on a specific domain or task.
For example, instead of a 175-billion-parameter general model, an SLM might have only a few hundred million parameters targeted at, say, legal document summarization or customer service chat. These models can be trained in minutes or hours instead of days, even on modest hardware. The key advantages are efficiency and control.
SLMs are more efficient and economical, using less memory and energy to run, which makes them deployable on local servers, PCs, or even smartphones. This means companies can train and run their own AI models tailored to their data and needs, rather than sending data to third-party APIs. It reduces dependency on tech giants and addresses data privacy concerns by keeping sensitive data in-house. It also democratizes AI—smaller firms or research labs with limited budgets can develop useful AI without needing a billion-dollar infrastructure.
Why It Matters:
This shift could lead to a proliferation of custom AI models across industries. Instead of one-size-fits-all large models, we may have a landscape of many specialized AIs—for medical records, for engineering design, for local government use, etc. Such specialization can improve accuracy and relevance in each domain, since the model isn’t cluttered with “knowledge” it doesn’t need.
Moreover, SLMs often have fewer issues with “black-box” unpredictability; they can be easier to interpret and debug than gargantuan models, making it feasible to ensure fairness and reduce biases in specific use cases.
Importantly, focusing on small models addresses the sustainability issue of AI. Training giant LLMs consumes vast energy and resources (training GPT-4 reportedly cost over $100 million and tons of electricity). SLMs are far less resource-intensive, aligning with calls for more eco-friendly AI development.
Recent Developments:
Over the past year, several open-source SLMs and tools have been released. For instance, Meta’s LLaMA model (2023) showed that a 7B-parameter model, when fine-tuned, could approach the performance of much larger models on many tasks, sparking a wave of innovation in the community.
Startups like MosaicML and Alpaca have produced recipe and code for training smaller models cheaply. Even Google has signaled interest: it introduced a series of smaller models (e.g., Gemini Mini) aimed at being more efficient.
Additionally, new techniques like knowledge distillation (compressing a large model’s knowledge into a small model) and low-rank adaptation have matured, making it easier to create high-performing SLMs quickly.
In practical use, companies are already deploying SLMs: for example, hospitals fine-tuning small models on medical texts to assist doctors with diagnoses, or banks using small models internally for document classification without sending data to outside servers.
The debate around AI size has intensified at AI conferences—many researchers argue that after a point of diminishing returns, it’s more valuable to improve algorithms or data quality than to just scale up parameter count.
Even regulators and governments are exploring SLMs, as they offer a path for nations to have their own AI systems (trained on local languages or norms) without needing the infrastructure of an OpenAI.
Challenges and Concerns:
Small models face a few challenges. By definition, an SLM may not match the broad capabilities of a giant model—it might excel at its narrow task but lack flexibility.
There is a risk of fragmentation: if everyone trains their own small model without the rigorous testing that big models undergo, quality could vary and some models might have hidden biases or vulnerabilities.
Also, while SLMs reduce dependency on tech giants, not everyone has the expertise to train them, which could create a different kind of digital divide (companies with AI talent vs. those without).
Another concern is that if not carefully managed, privacy could be at risk—ironically, having many small models floating around might mean more vectors for leaks or misuse, unless security is strong.
From a competitive standpoint, large-model providers are not standing still; they might offer smaller fine-tuned versions of their big models as products, potentially crowding out open-source efforts.
There’s also an ongoing scientific question: How small is too small?
We don’t yet fully understand the limits. A model that’s too small might simply not capture the complexity of language or the task. Researchers are actively studying how to measure an SLM’s understanding and ensure it’s not missing crucial context.
Despite these challenges, the momentum behind efficient AI is real. As one analyst put it:
“The era of simply chasing bigger model sizes is ending; the future is about doing more with less.”
This means we can expect smarter optimization, hybrid systems (combining small models), and a focus on “right-sized” AI that balances power and efficiency.
4. Cattle “Burping” Remedies (Methane Reduction)
Overview: Surprisingly, one of the biggest climate change culprits is the humble cow. Livestock farming, especially cattle, produces enormous amounts of methane through the natural digestion (enteric fermentation) process—primarily via cow burps.
Methane is a greenhouse gas dozens of times more potent than CO₂ in the short term, and livestock emissions account for an estimated 12–20% of global greenhouse gases (mostly methane). In fact, cattle alone are often cited as responsible for roughly one-third of human-caused methane emissions.
To tackle this, researchers and startups have been developing “cattle burping remedies”—innovative solutions to reduce methane produced by cows. These range from dietary additives (like special feed ingredients that inhibit methane) to wearable gadgets for cows that neutralize methane.
Reducing emissions from cows is seen as a critical and relatively quick win for climate action, since cutting methane has almost immediate benefits for slowing warming.
How It Works and Why It Matters:
The various remedies target the source of methane, which is produced by microbes in the cow’s stomach as they digest fibrous food. Feed additives are a leading approach. By supplementing cattle feed with certain compounds or plants, scientists can suppress the methane-producing microbes or alter the fermentation process in the rumen.
For example, an additive called 3-NOP (commercially known as Bovaer) inhibits an enzyme involved in methane synthesis. Feeding just a quarter teaspoon of 3-NOP daily to a cow can cut methane emissions by about 30% for dairy cows and up to 45% in beef cattle. Another approach uses a natural solution: red seaweed (Asparagopsis taxiformis). When a small amount of this seaweed is added to cattle feed, it can disrupt methane production dramatically—studies have shown reductions of around 80% in methane output. This is because compounds in the seaweed (bromoforms) chemically block methane formation.
Beyond diet, a novel tech solution is the methane-capture mask. A startup called Zelp has created a facemask for cows that fits over their nostrils and oxidizes methane as the cow exhales. Inside the device, methane is converted into CO₂ and water vapor by a catalyst, reducing its climate impact (since CO₂, while still a greenhouse gas, is far less potent per molecule than methane). Zelp’s mask can eliminate about 60% of methane emissions per cow when worn continuously. Such technology is significant because it doesn’t require changing what the cow eats—it directly treats the emissions.
Why It Matters:
Reducing methane from cows is highly impactful for climate action. Methane has over 80 times the warming potential of CO₂ over a 20-year period, so cutting it can slow global warming faster than almost any other measure. Unlike decarbonizing sectors like transportation or power, which require big infrastructure changes, mitigating cow emissions can be done relatively quickly by changing farming practices.
It’s estimated that widespread adoption of these remedies could significantly reduce global greenhouse gas emissions. Additionally, these solutions are often safe for the animals and can even improve feed efficiency. Some farmers report that cows actually grow slightly better on the additives since less energy is lost as methane. This means farmers have an incentive—reduced emissions and potentially better productivity. With the global cattle population over 1.5 billion, implementing these fixes at scale could be a game-changer in meeting climate targets.
Recent Developments:
Over the last couple of years, there have been promising trials and moves toward commercialization. 3-NOP (Bovaer), developed by DSM, has received regulatory approvals in places like the EU, Brazil, and Australia for use in cattle feed. Large dairy companies have started pilot projects feeding Bovaer to herds to validate the 30% emission reductions in real farm conditions.
Seaweed supplements have moved from lab to farm trials. Researchers in Australia and California conducted longer-term studies showing sustained methane reductions (around 80%) in cattle fed small daily doses of dried red seaweed. Challenges like sourcing enough seaweed and ensuring cost-effectiveness are being addressed, including the idea of farming seaweed at scale specifically for this purpose.
Meanwhile, the Zelp methane mask gained attention by winning awards (even the UK’s Prince Charles lauded it), and it’s been tested on farms in Argentina and Europe. The company is refining the design for comfort and efficacy, aiming for a 60% reduction mark.
Another interesting development is in breeding programs. Some research suggests certain cows naturally emit less methane, raising the possibility of selectively breeding low-methane cattle over time. Governments and industry groups are increasingly on board—New Zealand, for example, is investing in research and may become one of the first countries to regulate agricultural emissions, encouraging farmers to adopt these tools.
Notably, the European Union’s climate plan and other international efforts highlight livestock methane reduction as a key strategy. We’ve also seen the first carbon credit methodologies for reduced methane, meaning farmers who cut cow emissions might earn carbon credits to sell, creating financial rewards. This year at SXSW, discussions highlighted that multiple approaches might be combined—e.g., a cow could be fed an additive and wear a methane-capturing device for maximum effect.
Examples of Methane Reduction Solutions for Cattle:
| Solution | Method | Methane Reduction |
|---|---|---|
| Red Seaweed | Add small amount of Asparagopsis seaweed to diet. Compounds in seaweed inhibit methane-producing microbes. | Up to ~80% reduction |
| 3-NOP (Bovaer) | Add synthetic 3-nitrooxypropanol to feed. Inhibits an enzyme in the methane formation pathway. | ~30% reduction (dairy cows); up to ~45% in beef cattle. |
| Methane-Capture Mask (Zelp) | Fit a catalytic mask over cow’s nostrils. Oxidizes methane in exhaled breath to CO₂ and H₂O. | ~53–60% reduction (with continuous use). |
Key metrics from trials indicate significant emission cuts. Combining measures (diet + tech) could potentially achieve over 80% methane reduction per animal.
Challenges and Concerns:
While these solutions are promising, implementing them at scale presents challenges. Cost and practicality are major factors—convincing millions of ranchers globally to adopt feed additives or mask devices is no small task. Feed additives like 3-NOP and seaweed need supply chains to produce them in bulk; seaweed, in particular, would require large-scale aquaculture or natural harvesting, raising environmental and logistical questions. The additives must also be affordable for farmers or subsidized; otherwise, uptake will be limited.
For the methane-capture masks, managing herds with wearable devices is labor-intensive. Cows would need to wear them all day, so ensuring the devices stay on, remain effective, and do not harm animal welfare is critical. Early tests claim no adverse effects on cow behavior or stress levels, but farmers will need to be convinced.
Another challenge is adoption in developing countries, where much of the world’s cattle population resides. Small-scale farmers may not have access to these technologies or the means to apply them. Policy support is crucial—if governments incentivize methane reduction through carbon credits, subsidies, or mandates, adoption could accelerate.
There’s also the question of long-term effectiveness. Cows are complex biological systems; some fear that microbes could adapt over time to additives, or that effectiveness might wane. Ongoing research is monitoring whether methane production rebounds over longer periods.
Additionally, as one farmer humorously noted, “What about the other end of the cow?” While most methane comes from burps, manure management is also needed to capture methane from waste. Solutions like anaerobic digesters can help turn manure emissions into usable biogas.
Finally, consumer and cultural factors play a role. Farmers take pride in natural, healthy herds—any hint that an additive might affect milk or meat quality must be addressed with evidence. So far, studies show no negative impact, and milk from cows on Bovaer has been deemed safe to drink.
In summary, Cattle methane remedies are a bright spot in climate innovation—a mix of biotech and clever engineering that could deliver fast results. The challenge is ensuring these solutions are scalable, economically viable, and widely adopted. If they are, the planet stands to benefit enormously from a significant dent in greenhouse emissions.
5. Robo-Taxis
Figure: A prototype Waymo fully autonomous robo-taxi (displayed at a museum) – a custom electric vehicle laden with sensors (lidar dome on top, cameras and radar) but no steering wheel. Companies like Waymo, Cruise, and others are racing to deploy similar vehicles for ride-hailing services on public streets.
Overview: Self-driving taxis – or robo-taxis – have been a dream for over a decade, and now that dream is closer than ever to reality. These are vehicles that ferry passengers without a human driver at the wheel. In 2025, we stand at a tipping point: autonomous ride-hailing services are now operating in several cities, and companies are planning broader rollouts. Waymo (Google/Alphabet’s self-driving unit) and GM’s Cruise have been offering driverless rides in cities like San Francisco and Phoenix.
At SXSW, Niall Firth noted that after years of hype and setbacks, “an old dream is moving nearer” to everyday life. Robo-taxis promise to improve road safety (by eliminating human error) and provide convenient mobility on demand, potentially at lower cost than traditional taxis (since no driver wage). They are a flagship technology for AI and robotics, combining advances in computer vision, machine learning, sensors (lidar, radar, cameras), and real-time decision-making.
How It Works and Why It Matters:
A robo-taxi uses an array of sensors to perceive its environment – detecting other cars, pedestrians, traffic signals, and more – and onboard AI software to make driving decisions. High-definition maps and powerful computing guide the vehicle through city streets.
Why it matters: Safety and efficiency. In theory, autonomous cars don’t get distracted, tired, or drunk; they react faster than humans and can communicate with each other to smooth traffic flow. If perfected, they could drastically reduce accidents (which today claim over a million lives globally each year). Additionally, robo-taxis can be summoned on demand via apps, potentially reducing the need for personal car ownership in cities – this could lower congestion and parking woes if fewer people drive their own cars and use shared autonomous rides instead. For those who cannot drive (elderly, disabled, etc.), robo-taxis offer new independence.
Economically, the impact is huge: transportation and taxi services could be revolutionized. Companies like Uber and Lyft have long-term plans that hinge on autonomous vehicles to cut costs. The technology also has societal implications – for instance, professional drivers (taxi, rideshare, truck drivers) could be gradually displaced, while new jobs in overseeing or maintaining autonomous fleets might emerge. This technology is also a proving ground for AI in the real world: success in robo-taxis would accelerate autonomous tech in delivery, trucking, and even things like autonomous wheelchair or delivery robots.
Recent Developments:
In the last couple of years, robo-taxis have moved from testing to initial deployment. In San Francisco, Cruise and Waymo both received permits to operate fully driverless cars (no safety driver) commercially in parts of the city. As of late 2023, Waymo announced it was providing over 150,000 rides a week in its serviced areas. The service areas have been expanding – Waymo is expanding to Austin, TX (after Phoenix, SF, and Los Angeles), even partnering with Uber to offer robo-taxi rides there. Meanwhile, in China, companies like WeRide, Pony.ai, and Baidu’s Apollo have made big strides: at least 19 cities in China are running robotaxi trials, and Pony.ai plans to scale up to 1,000 driverless vehicles in its fleet. One of the first truly driverless services (no human in car at all) opened to the public in Beijing and Wuhan, showing China’s push in this field. This global race has led to major investments – for instance, GM poured $10+ billion into Cruise, and startups raised huge funds as well (e.g., WeRide’s multi-billion valuation).
However, it hasn’t all been smooth. Cruise experienced some highly publicized incidents in San Francisco – their cars occasionally stopped in traffic unexpectedly or failed to yield to emergency vehicles, causing public concern. Indeed, in late 2024, California regulators suspended Cruise’s permit after a series of incidents, leading GM to halt all Cruise operations nationwide pending investigations. This was a reality check that even after millions of test miles, the technology isn’t foolproof. Public acceptance is still being earned: some SF residents even staged protests (placing cones on robotaxi hoods to disable them) to express frustration at the vehicles’ quirks. Tesla, which has a different approach (selling a “Full Self-Driving” assist feature to private car owners), announced plans to deploy its own robotaxi services in Texas and California, though their tech is controversially still Level 2 (requiring human oversight) and not yet true autonomy. At SXSW, it was noted that Waymo’s expansion to Austin and other cities signifies the beginning of broader adoption, but “whether society is ready to remove human drivers entirely remains open.”
In short, recent developments show rapid progress, tempered by growing pains. The first commercial robo-taxi services are here, scaling up city by city, and regulators are feeling their way around how to oversee them.
Challenges and Concerns:
The robo-taxi revolution faces several hurdles:
- Safety and Public Trust: Any accident or mistake by a driverless car grabs headlines, and rightly so – companies must prove they are at least as safe as a good human driver. Edge cases (unusual scenarios) can still confuse AI: e.g., unexpected road construction, a police officer directing traffic (non-standard signals), or pedestrians behaving erratically. Building AI that handles every scenario is extremely hard. When mistakes happen, like a Cruise car blocking an ambulance as alleged in one incident, it erodes trust. Gaining public trust will require transparency (sharing safety data) and consistent performance.
- Regulation: Different cities and countries have different rules. Navigating regulatory approval is complex – some jurisdictions are cautious after seeing issues in SF, while others (like parts of China or Phoenix, AZ) are more permissive. Regulators also worry about cybersecurity (could a robo-taxi be hacked and turned into a weapon?) and liability (who is responsible in a crash – the company, the passenger, the manufacturer?). These legal frameworks are still being ironed out.
- Ethical and Economic Impact: If robo-taxis proliferate, what happens to the livelihoods of millions of professional drivers globally? There are concerns about job losses. On the other hand, there’s an argument that it could fill labor shortages in areas like logistics and that new tech jobs will be created. Ethically, programming decisions (like how to react in an unavoidable crash scenario – the so-called trolley problem) come into play, and companies have to consider public input on such matters.
- Technical Challenges: Weather and geography can be tricky – heavy snow or dense fog can interfere with sensors. So areas with mild climates (like Arizona) saw earlier deployments, whereas places with winter weather are next to tackle. Also, scaling up from geofenced urban cores to broader areas (suburbs, highways) is an ongoing technical hurdle.
- Social Acceptance: Some communities have reacted negatively. In San Francisco, aside from protests, there have been reports of emergency services frustration with robo-taxis blocking fire scenes, etc. Cities will need to adapt protocols (for instance, how first responders can remotely disable or move an autonomous vehicle if needed). Additionally, some people are simply uncomfortable without a human at the wheel, at least until the technology’s safety is well proven.
Despite these challenges, the momentum seems irreversible. As one industry observer said, “the question is no longer if robo-taxis will come, but when and how.” The coming 2-3 years will likely see a patchwork: some cities with bustling robo-taxi networks, and others holding off. By learning from each, best practices will evolve. The long-term potential – safer roads, less congestion, mobility for all – means society has a strong incentive to solve the challenges and embrace the robo-taxi revolution carefully.
6. Cleaner Jet Fuel (Sustainable Aviation Fuel)
Overview: The aviation industry is notoriously difficult to decarbonize – jet aircraft typically burn kerosene (a fossil fuel) and produce significant CO₂ and other emissions at high altitude. In fact, if global aviation were a country, it would rank in the top 10 emitters of CO₂.
Enter Cleaner Jet Fuel, specifically Sustainable Aviation Fuel (SAF), as a breakthrough to tackle this problem. SAF refers to non-petroleum-based fuels (like biofuels made from plant oils, algae, or waste, and synthetic fuels made from captured CO₂ + green hydrogen) that can be used in place of conventional jet fuel but with a dramatically lower net carbon footprint. The push for SAF has gained urgency: decarbonizing air travel remains one of the biggest unsolved climate challenges.
At SXSW 2025, this was highlighted as a critical technology – one that could transform aviation without waiting for futuristic hydrogen or electric planes to become viable for long-haul flights.
How It Works and Why It Matters: Sustainable Aviation Fuels come in a few varieties.
- Biofuels: made from biological sources like used cooking oil, plant biomass, or even municipal waste. These are refined to produce a fuel that is chemically similar to jet kerosene and can be dropped into existing aircraft engines. Because the carbon in biofuels was absorbed from the atmosphere by plants recently, burning it just returns that carbon, resulting in a much smaller net addition than fossil fuel (which releases carbon locked away for millennia).
- E-fuels (electrofuels): made by taking carbon dioxide (captured from the air or an industrial source) and combining it with hydrogen produced by renewable electricity (via electrolysis of water) to create a synthetic hydrocarbon fuel. This process essentially recycles CO₂ into fuel. Companies like Twelve are doing exactly this – using CO₂ as feedstock to create jet fuel. If the required energy comes from renewables, the fuel can be nearly carbon-neutral (the CO₂ emitted in flight was the CO₂ taken from air to make it).
Why it matters:
Using SAF can cut aviation’s carbon emissions by 70-100% on a life-cycle basis (depending on the source and production method) compared to conventional fuel. This is huge for climate goals – there are no easy alternatives for long-distance air travel on the near horizon. Batteries are too heavy for most large aircraft, and hydrogen would require new plane designs and infrastructure. SAF, however, works with current planes and engines (most SAF blends are certified up to 50% with regular fuel now, and the goal is 100% SAF flights in coming years). Thus, it’s a drop-in solution that can start reducing emissions immediately. Additionally, SAF often results in less sulfur and particulate pollution, improving air quality around airports.
Recent Developments:
In 2023 and 2024, we saw significant moves: Production of SAF is ramping up, albeit from a very low base. Major airlines have started signing multi-year offtake agreements with SAF producers. For instance, United, Delta, and others have deals to buy millions of gallons of SAF from various biofuel refineries. A key policy boost came from the European Union – ReFuelEU initiative, which mandates that airlines refueling at EU airports must uplift a certain percentage of SAF: starting at 2% in 2025 and rising steeply to 70% SAF by 2050. This regulatory push provides certainty to producers that demand will be there, encouraging investment in new SAF plants. Similarly, tax credits for SAF were introduced in the U.S. (e.g., in the Inflation Reduction Act) to make it more price-competitive.
On the innovation front, several companies reached milestones: LanzaJet, a spinoff from LanzaTech, began construction of one of the first alcohol-to-jet SAF production plants, using ethanol from waste sources to make jet fuel. Twelve (mentioned in the SXSW talk) produced demonstration quantities of jet fuel from recycled CO₂, and in one notable test, the U.S. Air Force flew a plane using fuel from CO₂. Startup Prometheus is another working on direct electrofuels. Also, some airlines undertook demonstration flights – e.g., one engine on a passenger flight running 100% SAF – showing it can perform similarly to normal fuel. Boeing and Airbus are both supporting efforts to certify aircraft on 100% SAF in the near future.
One exciting development: power-to-liquid projects in Europe (like in Norway and Germany) are planning to produce synthetic jet fuel using renewable energy and CO₂ from air. And H2 Green Steel (from the Green Steel topic) is not directly SAF, but note: the processes share a hydrogen focus – cheap green hydrogen is a key enabler for e-fuels. If green hydrogen projects succeed, e-fuel costs will drop.
However, despite these developments, current SAF supply meets only a tiny fraction (<1%) of aviation’s needs. According to industry data, about 1.9 billion liters of SAF might be produced in 2024, which is only ~0.5% of jet fuel demand. The good news is this is triple previous output, but the gap is enormous.
Challenges and Concerns:
The number one issue is scaling up production and bringing down cost. Today, SAF can cost 2–5 times more than conventional jet fuel, depending on feedstock and method. This makes airlines hesitant to use it in large quantities without policy support (as fuel is a major operating cost).
Feedstock availability is another constraint: there is a limit to used cooking oil or agricultural waste that can be easily collected – relying solely on biofuels could compete with food production or cause land-use issues if not managed carefully. There are sustainability criteria to ensure SAF truly reduces emissions and doesn’t create other problems (for example, biofuel from palm oil could be worse if it drives deforestation – so that’s typically excluded from “sustainable” definitions).
Production capacity:
Building new biorefineries or electrofuel plants takes time and billions in investment. There’s a risk that these investments might falter if policies change or if a future breakthrough (like hydrogen planes) makes SAF less needed – though most experts think we’ll need SAF through mid-century at least.
Airlines also face the logistical challenge of distributing SAF:
Initially, SAF is being delivered to major hubs and blended into the fuel supply. But getting it to all airports worldwide requires a fuel logistics network build-out.
Technical:
While current engines can use up to 50% SAF (blended with regular fuel) with no modification, going 100% SAF may need some tweaks (mainly because SAF can lack aromatics that help seal swelling in fuel systems – engineers are working on that).
Economic and Policy Concerns:
If SAF remains expensive, it could drive up ticket prices, potentially reducing demand for flying or making air travel more exclusive. Some have even posited that without cheap SAF, the era of abundant cheap flights might wane (though that’s speculative). However, given the climate imperative, many argue that’s a necessary adjustment – the alternative of continuing high emissions is unacceptable. The aviation sector has set goals (like net-zero by 2050) that heavily rely on SAF, so there’s pressure to sort these issues.
Another concern is that focusing on SAF might delay other innovations (like electrification of short flights, or efficiency improvements to planes) – but realistically, the industry is pursuing all angles in parallel.
In summary, cleaner jet fuel via SAF is both essential and challenging. The SXSW discussion likely emphasized the “dilemma of aviation” – we need to fly (for global business, connection, etc.), but we also need to drastically cut emissions. SAF is the bridge to reconcile that dilemma. The coming years will be critical to determine if SAF can scale fast enough. With strong mandates (like the EU’s) and technological progress, there is cautious optimism. By 2030, we might see a noticeable fraction of flights powered by SAF, and by 2050, if all goes well, jet planes could be flying on fuel that’s mostly green, fulfilling the vision of cleaner jet fuel enabling sustainable air travel.
7. Versatile AI-Powered Robots
Overview: We’ve long imagined robots as multi-purpose helpers – think of a humanoid robot that can cook, clean, fetch objects, or work alongside humans in a factory on varied tasks. Historically, however, robots have been highly specialized and inflexible: an assembly-line robot might weld the same joint over and over, but it can’t suddenly switch to a different task.
Now, a new wave of AI-powered robots promises much greater versatility. By integrating generative AI and advanced machine learning, robots are learning to adapt to new tasks on the fly, guided by natural language or visual cues. This was described as “a new attempt” to finally achieve more human-like flexibility in robotics. Companies like Figure AI (a Silicon Valley startup) are developing humanoid robots equipped with powerful AI brains, aiming for use-cases from household chores to eldercare to warehouse work.
The convergence of AI and robotics is giving us machines that can perceive their environment, understand spoken or written instructions, and even learn new skills by example or through simulation – a significant leap from the single-task robots of the past.
How It Works and Why It Matters:
The key enabling technology here is the integration of Generative AI (particularly large language models and vision models) with robotic control. Generative AI provides a form of high-level “cognitive” ability – the robot can understand context, reason in natural language, and generate plans or sequences of actions. For example, you could tell a robot, “I spilled juice on the floor, please clean it up,” and the robot’s AI brain would interpret that, figure out it needs to fetch a mop or paper towel, go to the spill, and wipe it. This is incredibly hard for traditional robotics, but with language models and training, robots can now parse such requests. They use visual AI to recognize objects and situations (like locating the juice spill and a towel).They use reinforcement learning or motion planning to physically execute the task (moving their arms and navigating). Generative AI can even help the robot handle novel objects by extrapolating from what it knows – for instance, if the robot has never seen a particular brand of juice carton, its vision AI still detects “this is a carton, likely containing liquid” and handles it accordingly.
This matters because it potentially unlocks a huge range of applications. In industry, a versatile robot could be re-tasked with new assignments simply by telling it what to do – reducing programming costs and increasing productivity. In homes or healthcare, a single robot could assist with many activities: cleaning, cooking simple meals, helping someone out of bed, retrieving dropped items, etc. That’s revolutionary for aging societies where caretaker shortages are an issue. Moreover, versatility addresses a limitation: currently, deploying robots is often only cost-justified if the task is extremely repetitive. But if one robot can do 10 different things, it becomes far more valuable. It could also work safely in dynamic human environments (since AI helps it understand context and social cues).
Recent Developments:
A lot has been happening at the intersection of AI and robotics. Figure AI, for example, recently unveiled its “Figure 02” humanoid robot (an evolution of their first model) which integrates a dual AI system: one for high-level reasoning (a language model that interprets commands) and one for real-time motor control. In a demo, two Figure robots were shown collaborating on a task – putting away groceries: one robot picked items from a table and handed them to the other to place in a fridge. Impressively, these robots had not been explicitly programmed for that exact sequence; instead, the AI planned and executed it based on general knowledge of household tasks. Figure has also been testing their robot in an automotive factory (BMW) to see how it can learn various assembly tasks by observation and practice.
Another example: Google’s Robotics team combined their language model (PaLM) with a robot arm, enabling it to understand human instructions like “please clean up this spill” and then orchestrate a series of actions to do so. This project, called PaLM-SayCan, showed the robot deciding which motions to take, guided by the AI’s understanding of the goal. Likewise, Boston Dynamics (known for their agile robots like Atlas) have been quietly adding more smarts to their machines to go beyond pre-scripted parkour routines and do useful tasks. Generative AI tools have also helped in simulation and training – companies use virtual environments to let the AI “practice” countless scenarios (like picking up varied objects) so that the robot improves rapidly. NVIDIA’s Isaac Sim is one such platform widely used (and mentioned by Figure for training their models faster in simulation).
The SXSW mention highlighted Figure AI working on humanoid robots that respond to visual and auditory signals – essentially, robots that can see and hear like we do, and interpret those inputs intelligently. The “potential is enormous” as noted, from home to eldercare to industry. This year, we also saw other startups like Tesla (with their Optimus robot prototype) and Apptronik (with a general-purpose robot called Apollo) making strides. While Tesla’s robot is still early, they did demonstrate it sorting objects and responding to commands in a lab.
Challenges and Concerns:
Despite the excitement, we must ask: How far are we from a Rosie-the-Robot that can truly do it all? The SXSW commentary itself was cautious: “it remains questionable how far the technology really is.” Indeed, current demos, while impressive, are often tightly controlled. Robots struggle with unstructured environments – our world is chaotic, and robots can be thrown off by small changes (say, a command phrased in an unusual way, or an object that falls to the floor unexpectedly). Ensuring reliability and safety is a big challenge. A robot helping an elderly person must not accidentally knock them over or make a mistake carrying hot soup. That requires extensive testing and fail-safes.
AI Limitations:
The generative AI these robots rely on can sometimes misinterpret or produce incorrect actions (imagine an AI misunderstanding “put the milk on the table” and placing it on the edge where it falls). They also lack common sense knowledge at times (what if the spill is water vs. oil – how should cleanup differ? Humans know this intuitively, robots might not unless trained). Efforts are underway to imbue robots with more commonsense reasoning.
Physical Limitations:
Building a humanoid or multi-purpose robot is physically complex. Balancing, dexterous manipulation (like folding laundry), and safe operation around humans are engineering hurdles not completely solved. Battery life is another – advanced robots can be power-hungry, limiting how long they can operate untethered.
Cost:
These cutting-edge robots are expensive to prototype. Commercial viability will depend on cost coming down with mass production. If a versatile robot costs as much as a luxury car, adoption will be slow. However, many companies aim to eventually make them affordable, akin to the cost of a car or even less with scale.
Social and Ethical:
If robots become capable in workplaces, we could see displacement of some jobs (similar to concerns with automation in factories). But conversely, in sectors like eldercare where help is scarce, robots could fill crucial gaps. Ethically, ensuring robots behave in accordance with human values is important – they will be interacting directly with people, potentially in vulnerable settings. Privacy is a concern too: a home robot sees and hears a lot; robust safeguards will be needed so that sensitive data isn’t misused.
There’s also the yuck factor or fear factor – some people find humanoid robots unsettling (the “uncanny valley” if they look too human-like, or general apprehension of AI autonomy).Building trust through transparent operation (maybe obvious indicators of what the robot is doing or thinking) could help.
In conclusion, versatile AI-powered robots are on the horizon, “a new attempt” to achieve what sci-fi has long envisioned . They have immense potential across daily life and industry by adapting to tasks and environments. Yet, it’s early days. As these prototypes move to real-world pilots, we’ll learn where they excel and where they need improvement. It’s very plausible that in a decade, having a helpful robot at work or even at home will be as normal as having a smartphone – but to get there, technologists must surmount technical, economic, and social challenges. For now, the progress is exciting but we should temper expectations: as Niall Firth mused, robots have historically been “surprisingly inflexible” – this generation aims to change that, and if successful, it “could nonetheless change the world” by finally bringing versatile robotics from fiction to reality.
8. Long-Acting HIV Prevention Medication
Overview: After decades of efforts to curb the HIV/AIDS epidemic, a major breakthrough is here: a long-acting injectable medication that can prevent HIV infection. One such drug, Lenacapavir, was highlighted at SXSW as a game-changer in HIV prevention.
Unlike daily pills (like PrEP with Truvada) that require strict adherence, lenacapavir is designed to be given as an injection only twice a year. This means just one shot every six months can provide continuous protection against HIV. This is monumental for controlling HIV spread, especially in populations where taking a daily pill is challenging due to stigma, forgetfulness, or limited healthcare access.
The context is crucial: despite all we know about HIV, over 1 million people still acquire HIV each year worldwide. A long-acting preventive option could dramatically reduce new infections and help end the epidemic.
How It Works and Why It Matters:
Lenacapavir is a type of antiretroviral drug (specifically, a capsid inhibitor) that interferes with the HIV virus’s ability to replicate. When injected subcutaneously (under the skin), it forms a depot that slowly releases the drug over many weeks, maintaining effective levels in the body for six months. This long action is due to the drug’s chemical design – it’s slowly metabolized and stays in tissues that are initial sites of infection (like certain immune cells).
Why it matters:
The simplified dosing is a game-changer for adherence. Traditional PrEP (pre-exposure prophylaxis) with daily pills can be highly effective, but many struggle with daily regimens, and missing pills diminishes protection. With an injection every six months, adherence could be much higher – you only need to remember two clinic visits a year. This is particularly beneficial in regions with limited healthcare infrastructure: for example, a public health campaign could line people up twice a year for injections, rather than ensure a continuous pill supply. It also mitigates the stigma or privacy issues of taking a daily pill (some might hide pills or skip them to avoid others knowing; an injection is private and infrequent).
Moreover, long-acting PrEP reaches groups like young people or marginalized communities who might not engage with the healthcare system regularly – the low-barrier nature of it is key. Preventing HIV has individual and societal benefits: it saves lives, reduces medical costs (treating HIV is lifelong and expensive), and moves us closer to the elusive goal of zero new infections.
Recent Developments:
The excitement around lenacapavir comes from clinical trial results that have been nothing short of astounding. In a major trial called PURPOSE 1 involving women in sub-Saharan Africa, those who received lenacapavir injections had zero new HIV infections, compared to several in the placebo group – effectively 100% efficacy in preventing HIV among those who got the shot. This was reported in mid-2024, and it caused a stir in the global health community. To put it plainly, not a single woman who got the injection contracted HIV over the study period. A parallel trial (PURPOSE 2) in other populations (including men who have sex with men and transgender individuals) showed around a 96% reduction in incidence compared to expected background rates, also extremely high. These results led experts to call lenacapavir “potentially an important new tool” for HIV prevention.
Regulatory-wise, lenacapavir (branded Sunlenca) was already approved in some places for treating HIV (in combination therapy for patients with multi-drug-resistant HIV). Now, Gilead (the manufacturer) is seeking approvals for its use as PrEP (preventive). The FDA and other regulators have granted fast-track designations. It’s expected that by 2025–2026, lenacapavir for PrEP will be approved in multiple countries if all goes well.
Additionally, access efforts are underway: in October 2024, Gilead announced licensing agreements with generic manufacturers in 120 low-income countries to produce lenacapavir cheaply. This is crucial to avoid the mistakes of the early AIDS era when treatments took years to reach developing nations. With these licenses, once approved, generic versions could be made available in Africa and Asia relatively quickly at low cost. Advocacy groups and global health agencies are pushing hard on this front because the initial pricing of lenacapavir in wealthy markets is high (in the US, the treatment version costs tens of thousands of dollars per year – not feasible for prevention in poor regions). However, analyses by public health experts suggest it can be produced for as little as ~$40-$100 per person per year at scale, given it’s only two doses.
Challenges and Concerns:
There are a few challenges to address. Cost and access is primary: while Gilead’s licensing deals are promising, we need to ensure production ramps up and that distribution networks can deliver this biannual shot to the people who need it most. This includes potentially millions of individuals in high-prevalence areas – a logistics challenge for health systems. Funding will be needed (through international donors or domestic health programs) to subsidize or provide the injections for free to at-risk populations.
Another concern:
adherence to injection schedule. While two shots a year is infinitely easier than 365 pills, it still requires people to come for that shot. If someone delays or misses their 6-month injection, they could become vulnerable. Public health campaigns and reminders will be necessary. Interestingly, because lenacapavir levels taper slowly, even a delay of a few weeks might not immediately drop protection to zero, but it’s an area of active study how forgiving the schedule is.
Medical considerations:
Long-acting drugs raise issues: if someone inadvertently is HIV-positive (acute infection) when they get the shot, they’d essentially be on a single drug for treatment – HIV could develop resistance to that drug. Thus, proper HIV testing before each injection is important. Similarly, if someone contracts HIV despite or after the injection (rare but possible if timing goes wrong), the lingering drug could select for resistant virus. Fortunately, lenacapavir’s resistance profile and combination potential is known (as it’s also a treatment component), so mitigation strategies are in place (like adding another PrEP method if needed or ensuring follow-up testing).
Side effects and acceptance:
So far, lenacapavir has been well-tolerated; the main side effect is injection-site reactions (some people get bumps or irritation where the shot is given, since the drug sits under the skin). These have generally been mild. But acceptance will vary – some may be needle-averse. Education will be needed to explain the benefits. Also, unlike a pill you can stop and the drug leaves your system in days, once you get an injection, the drug stays for months. If someone has a bad reaction (unlikely, but say an allergy), there’s no quick way to remove it. Therefore, initial doses are typically monitored.
From a social perspective, long-acting PrEP could be a game-changer especially for women in parts of Africa. Many women can’t negotiate condom use or their partners object to pills – an injection every six months that the woman can keep private empowers them with protection. But some misinformation might arise (e.g., myths about fertility effects or such), so community education is important to ensure uptake.
Impact:
If rolled out widely, this innovation could bend the curve of new HIV infections significantly downward. UNAIDS and other organizations are extremely interested – they’ve called for making it widely available, noting estimates that generic production costs could be low. Combined with other measures (like traditional PrEP, condoms, and eventually an HIV vaccine if one is developed), it’s part of a multipronged strategy to end AIDS as a public health threat.
In summary, the long-acting HIV prevention injection is arguably one of the most exciting biomedical breakthroughs in recent years. It addresses a key vulnerability in prevention (adherence) and offers a practically 100% effective shield when used properly. The focus now shifts to ensuring it doesn’t become a luxury for the few but a lifesaver for the many. Efforts like Gilead’s generic licenses and international trials in diverse populations show promise. Overcoming cost, manufacturing, and last-mile delivery challenges will be crucial in the next 2-3 years. If successful, we might see HIV incidence plummet, bringing us closer to finally halting the 40-year scourge of HIV.
9. Green Steel
Overview: Steel is the backbone of modern civilization – used in buildings, cars, appliances, and infrastructure – but it comes with a huge carbon footprint.
The traditional process of making steel (using blast furnaces with coke/coal to reduce iron ore) emits large amounts of CO₂. In fact, steel production contributes roughly 7–9% of global greenhouse gas emissions, which is nearly triple the emissions of aviation. To meet climate goals, decarbonizing steel is essential.
Enter Green Steel – a term for steel produced with drastically lower (almost zero) carbon emissions. At SXSW, green steel was highlighted as a breakthrough, with examples like H2 Green Steel in Sweden and Boston Metal’s novel process. These developments aim to revolutionize an industry that has changed little in a century, cutting emissions by using hydrogen and clean electricity instead of coal. It’s not just an environmental effort; it has geopolitical and economic stakes too.
Countries or companies that master green steel could gain a competitive edge and upend the global steel trade (dominated by coal-heavy producers like China).
How It Works and Why It Matters:
Two main technological routes are leading the green steel push:
- Hydrogen Direct Reduction: This is the approach of H2 Green Steel and others (like the HYBRIT project in Sweden). Instead of using carbon (coke) to strip oxygen from iron ore, green hydrogen (H₂ produced via electrolysis using renewable energy) is used as the reducing agent. The reaction yields metallic iron (called direct reduced iron, or sponge iron) and water vapor (H₂O) instead of CO₂. The sponge iron is then melted in an electric arc furnace powered by renewable electricity to produce steel. When the hydrogen itself is produced with renewable power, the entire process can be nearly CO₂-free.
Why it matters: This could cut ~95% of emissions from steelmaking, which is huge. Also, using hydrogen ties steelmaking to the renewable energy sector – excess solar/wind can be used to make hydrogen, effectively storing energy in a usable commodity (steel). H2 Green Steel’s planned plant in Boden, Sweden exemplifies this: it will use local renewable energy (hydro and wind) to make hydrogen on-site and aims to produce 5 million tons of green steel annually by 2030. Such a plant would avoid millions of tons of CO₂ that a coal-based plant would emit.
- Electrolysis / Molten Oxide Electrolysis: This is the approach of Boston Metal (a U.S. startup). They are developing an inert anode electrolysis process (imagine something akin to how aluminum is produced, but for iron). Iron ore (iron oxide) is dissolved in a high-temperature electrolyte, and an electric current separates the iron and oxygen. The iron collects as molten metal, and oxygen gas is released – no CO₂ at all, assuming the electricity is clean. This is sometimes called molten oxide electrolysis (MOE). Boston Metal’s pilot is underway (with a larger pilot plant expected by 2024). This method directly produces liquid iron ready for steelmaking without coal.
Why it matters: It could be a more direct route that doesn’t even require hydrogen, just abundant clean electricity. It also can use low-grade iron ore that blast furnaces can’t, potentially broadening raw material options.
Green steel matters immensely for climate change – decarbonizing one of the hardest sectors. Additionally, from a market perspective: consumers (and automakers) are starting to ask for low-carbon materials. For example, Volvo and Mercedes have agreements to use green steel in car manufacturing by mid-decade. Companies that supply truly green steel could charge a premium and capture new markets as sustainability becomes a selling point.
Recent Developments:
In the past few years, the first green steel deliveries have happened. HYBRIT in Sweden produced a batch of steel using hydrogen reduction and delivered it to Volvo in 2021 (the “world’s first fossil-free steel”). H2 Green Steel raised over €1.8 billion and began construction of its large commercial plant in Boden. They plan to be operational by 2025-2026. Major steelmakers are not standing still either: ArcelorMittal, Thyssenkrupp, Salzgitter, and others in Europe are retrofitting plants to use hydrogen – many have targets to start hydrogen-based production before 2030.
In the U.S., Boston Metal has already built a pilot and is scaling up. They received investments from big players like Bill Gates’ Breakthrough Energy and steel giant ArcelorMittal. By 2024, they aim for a pilot-scale output, and they’re talking about commercialization later in the decade.
Challenges and Concerns:
- Cost: Green steel, at least initially, is more expensive. Using green hydrogen is costly because green hydrogen itself is pricey today (few dollars per kg). Estimates say green steel could be 20-50% more expensive to produce than conventional until technology matures and renewable energy gets even cheaper. This could make steel-intensive products (cars, buildings) more expensive if the difference is passed on. Over time, carbon prices and economies of scale might flip this – if carbon-heavy steel is penalized or as hydrogen costs drop, green steel might become competitive or even cheaper by the 2030s.
- Energy demand: Making steel with electricity (either via hydrogen electrolysis or direct electrolysis) requires a lot of power. Scaling green steel means scaling green electricity generation massively to supply that. For instance, H2 Green Steel’s plant will need several gigawatts of continuous power for its hydrogen production – Sweden can manage with hydro and new wind, but not every region has that capacity readily. Ensuring enough renewable energy and grid stability is a challenge.
- Hydrogen supply chain: We need infrastructure to produce, store, and deliver hydrogen if many steel plants convert. This is part of a broader hydrogen economy issue – pipelines, storage tanks, etc., need to be built.
- Technical scaling: Hydrogen-based direct reduction is already proven at pilot scale (companies like Midrex and others have run demonstration modules). But doing it at full industry scale and integrating with steel mills will have learning curves. Similarly, Boston Metal’s MOE process has never been done at industrial scale; they have to discover and solve issues as they scale (like electrode durability, maintaining purity, etc.).
- Raw materials: Green steel processes might need high-grade iron ore (for hydrogen DRI, the ore needs to be good quality or require beneficiation). There might be supply pinch-points if everyone shifts at once, until mining adjusts.
- Developing country producers: Places like India rely on cheaper, coal-based methods (often using coal directly in something called blast furnace-basic oxygen furnace or the coal-based DRI method). Transitioning their industries is a financial and technical challenge. There’s a risk that if only wealthy countries adopt green steel, others might become dumping grounds for cheap dirty steel. Global coordination (like CBAM or climate finance to help the transition) is needed to avoid simply shifting emissions around.
- Workforce and retrofitting: Existing steel plants (with huge sunk costs) might become stranded or need expensive overhauls. Managing this transition for workers and companies is important to avoid resistance. European steelmakers are already making plans, but some will close old blast furnaces, which affects communities.
- Potential Geopolitical tension: If countries with coal-heavy steel feel squeezed out (e.g. China or India facing tariffs), it could cause trade tensions. Conversely, if they get on board with green steel (China has some of the largest green H₂ projects announced), it could also be a collaborative win.
In conclusion, green steel is both a technological and industrial revolution in the making. As SXSW notes, it’s a fight not just about emissions but “emissions and influence” – meaning whoever leads in green steel could reshape the global economic landscape, currently dominated by old powers of coal and ore. The hope is that by 2030, multiple green steel plants will be operating, and by 2050, most new steel will be made without fossil fuels, drastically cutting one of the largest sources of carbon emissions. It’s an ambitious path, but one that appears increasingly feasible and indeed underway.
10. Effective Stem Cell Therapies
Overview: For decades, stem cell therapy has been touted as a potential cure-all – the idea of regenerating tissues or organs and treating diseases by using stem cells that can turn into any cell type needed.
After early hype in the 2000s, progress was slower than hoped, with many experimental treatments and a fair share of setbacks. But now, around 2025, we are finally seeing the first effective, clinically proven stem cell therapies for serious diseases emerge.
At SXSW, Niall Firth highlighted that “after decades of promise,” the first realistic clinical applications are arriving. Examples include treatments for Type 1 Diabetes and Epilepsy that have shown remarkable results in trials.
Essentially, scientists are learning how to grow the specific cells needed (like insulin-producing beta cells for diabetes, or inhibitory neurons for certain brain disorders) from stem cells and transplant them into patients, with life-changing outcomes.
How It Works and Why It Matters:
Stem cells are cells that can differentiate into other cell types. There are embryonic stem cells (pluripotent, from embryos) and induced pluripotent stem cells (iPSCs, reprogrammed from adult cells to behave like embryonic ones). Researchers can take iPSCs from a patient or a donor and coax them into becoming the specialized cells that the patient is lacking or that are malfunctioning.
For Type 1 Diabetes:
The disease is caused by the immune system destroying the pancreatic beta cells that produce insulin, leaving patients dependent on insulin injections. A stem cell therapy approach – used by companies like Vertex and ViaCyte – is to derive healthy pancreatic islet cells (including beta cells) from stem cells and transplant them into diabetic patients. Recent trials found that some patients who received such transplants were able to produce their own insulin again, effectively freeing them from daily insulin shots. In one case, a man with diabetes for 40 years got a stem cell-derived islet cell infusion and became insulin-independent for the first time – essentially a functional cure as long as the cells persist. This matters immensely: it hints that a one-time (or infrequent) cell therapy could replace constant medication and monitoring, giving better glucose control and preventing complications of diabetes. It validates decades of research and could apply to millions of people if made practical.
For Epilepsy (particularly severe drug-resistant epilepsy):
One experimental approach is to use stem cells to create inhibitory neurons (like GABAergic interneurons) and implant them into the brain. These new neurons can integrate and release inhibitory neurotransmitters to calm hyperactive neural circuits. A trial by a company (Neurona Therapeutics, referenced at SXSW) injected such cells into patients with refractory epilepsy. The result: seizures were dramatically reduced – over 90% fewer seizures in the first two patients a year after treatment. One patient was nearly seizure-free, a massive improvement for someone who might have had daily seizures prior. This is groundbreaking because it shows we can graft new neurons into the adult human brain to restore function – something that was science fiction not long ago.
These successes matter because they open the door to treating many conditions: not just diabetes and epilepsy, but potentially Parkinson’s disease (dopamine neurons from stem cells – trials are ongoing), heart failure (injecting stem-cell-derived heart muscle cells to repair heart tissue), macular degeneration (retinal pigment epithelial cells to restore vision – some trials have shown vision improvement), and more. It’s a validation that stem cell technology is reaching a maturity where real patients are benefitting.
Recent Developments: Several high-profile clinical trial results came out in the last couple of years:
- In the diabetes space, Vertex Pharmaceuticals announced in late 2021 that the first patient treated with their stem cell–derived islet cells (VX-880) had a stunning outcome: his body produced insulin at levels that allowed him to stop insulin injections entirely (with good blood sugar control). By 2023, Vertex reported two patients insulin-independent and others with major insulin reduction. This led them to a pivotal trial phase (though they had a temporary pause due to safety monitoring after two patients – likely a precaution in trials). Another company, ViaCyte, has also been testing implantable devices with stem cell islets. Mergers and collaborations (ViaCyte was acquired by Vertex) are happening to pool know-how.
- In the neurology space, beyond epilepsy, Parkinson’s disease trials with stem cells are underway. Early this year, a biotech reported initial safety and some hints of efficacy in transplanting dopamine neurons into Parkinson’s patients (though results are preliminary). Meanwhile, for spinal cord injury, some stem cell trials showed improved motor function in a few patients (though it’s still experimental). The SXSW mention of epilepsy likely referenced the Neurona trial or related work: by mid-2024, they reported >90% seizure reduction in initial subjects and no serious adverse effects, which is very promising.
- Also mentioned: epilepsy and Type 1 diabetes specifically, and that “first patients need no insulin anymore, epileptics have far fewer seizures,” which aligns with the above achievements.
- Another area: Sickle cell disease cure via stem cells – not exactly the same (that’s more gene therapy using bone marrow stem cells), but it’s in the realm of cell therapy and is now curing sickle cell in trials via gene-edited bone marrow transplants (though SXSW focus was more on regenerative tissue).
- In terms of regulatory approvals, a few stem-cell products have been approved: for instance, a therapy for severe corneal injury uses a patient’s own stem cells to regrow corneal tissue. And an approved therapy for a pediatric neurological disorder (CLN2 Batten disease) uses stem cells to deliver an enzyme. But those are niche. The ones discussed are broader.
Challenges and Concerns: Despite these successes, stem cell therapies still face significant challenges:
- Safety: Introducing cells into a patient’s body carries risks: tumors (if any undifferentiated cells remain, they could form teratomas), inappropriate integration, immune rejection (if not using patient’s own cells or an immune-protected device). In the diabetes trials, patients required immunosuppressive drugs to prevent rejection of the new islet cells (since those were derived from stem cells not from the patient’s own cells). Immunosuppression has its own risks. Researchers are working on “encapsulation” devices to protect transplanted islet cells from immune attack, or gene-editing the cells to be invisible to the immune system, to remove the need for immunosuppression. Until that’s solved, widespread use might be limited to severe cases where immunosuppression is worth it.
- Cost and complexity: These are not simple pills; they are living cell products. Manufacturing them consistently (the right cells, no contaminants) is complex and expensive. Vertex’s diabetes therapy, if approved, could be very costly (though one hopes competition and improved methods lower price). Healthcare systems might struggle with reimbursement, though one could argue a “cure” that frees someone from lifelong costs (of insulin, glucose monitors, hospitalizations for complications) could be cost-effective in the long run.
- Scaling up production: Each patient may need billions of cells. Scaling to treat millions of diabetics or others might require major biomanufacturing capacity. It’s a challenge to avoid batch-to-batch variability and to meet demand.
- Ethical/regulatory: Early on, stem cells (especially embryonic) had ethical debates, but now induced pluripotent cells ease that. Still, rigorous regulation is needed to ensure therapies are safe. In the past, unproven “stem cell clinics” popped up offering questionable treatments – regulatory bodies have been cracking down, but patient desperation still drives some to unapproved treatments. With real therapies emerging, it’s important to channel patients to legitimate trials or approved options, and stamp out charlatans.
- Efficacy and long-term durability: We have positive results in small numbers of patients with limited follow-up. Will these therapies remain effective for years? Will the cells survive long-term and continue functioning? Or will the diseases or immune responses catch up? For example, in type 1 diabetes, even if you put new beta cells, the autoimmune process could potentially attack them too if not modulated. Some approaches are co-transplanting cells with immune-protective measures. Long-term data will tell if these are cures or treatments that might need repeating or adjunct therapy.
- Price and equity: There’s a risk such treatments could be available only to the wealthy or in wealthy countries at first, given cost. Ensuring global access (like training centers to do stem cell transplants in developing countries) will be a challenge. Similar to how CAR-T cell therapies (for cancer) are amazing but extremely expensive and complex, we have to be careful stem cell therapies don’t widen health inequity.
Nevertheless, the momentum is building. The SXSW summary likely ended by noting that while not every promise has been realized, “first clinical applications finally seem realistic” and that some patients are seeing decisive benefits (no insulin, far fewer seizures) . This suggests that after many “next attempts” and iterations, we’re entering the era where stem cell therapies move from lab hype to hospital reality for certain conditions. It’s a big step in medicine – essentially fulfilling the dream of regenerative medicine by replacing or repairing tissues to cure diseases, rather than just treating symptoms.
10 Breakthrough Technologies 2025 – Highlights from SXSW
Presenting these ten breakthrough technologies at SXSW 2025, Niall Firth underscored an overarching theme: technological progress often happens in leaps after long gestation. Some innovations – like AI in search or autonomous cars – are already disrupting industries and society, raising urgent questions even as they solve problems. Others – like green steel or long-acting HIV meds – show how science and engineering can tackle the colossal challenges of climate change and public health when the will and resources are invested. And importantly, as Firth noted, not all promising ideas reach fruition; some will stumble due to economic or societal barriers . It’s a reminder that innovation must be coupled with practical feasibility and public acceptance. Yet, the lineup of breakthroughs from astronomy to medicine in 2025 suggests that what once seemed like science fiction isfast becoming reality. These technologies, if guided responsibly, could profoundly shape a more sustainable, healthy, and connected future, truly meriting the title “breakthrough”.
Fotocredits: Prompted & Crafted by onehundred.digital | Where Pixels Tell Stories
