In 2017, fresh off a PhD on theoretical chemistry, John Jumper heard rumors that Google DeepMind had moved on from building AI that played games with superhuman skill and was starting up a secret project to predict the structures of proteins. He applied for a job.

Just three years later, Jumper celebrated a stunning win that few had seen coming. With CEO Demis Hassabis, he had co-led the development of an AI system called AlphaFold 2 that was able to predict the structures of proteins to within the width of an atom, matching the accuracy of painstaking techniques used in the lab, and doing it many times faster—returning results in hours instead of months.

AlphaFold 2 had cracked a 50-year-old grand challenge in biology. “This is the reason I started DeepMind,” Hassabis told me a few years ago. “In fact, it’s why I’ve worked my whole career in AI.” In 2024, Jumper and Hassabis shared a Nobel Prize in chemistry.

It was five years ago this week that AlphaFold 2’s debut took scientists by surprise. Now that the hype has died down, what impact has AlphaFold really had? How are scientists using it? And what’s next? I talked to Jumper (as well as a few other scientists) to find out.

“It’s been an extraordinary five years,” Jumper says, laughing: “It’s hard to remember a time before I knew tremendous numbers of journalists.”

AlphaFold 2 was followed by AlphaFold Multimer, which could predict structures that contained more than one protein, and then AlphaFold 3, the fastest version yet. Google DeepMind also let AlphaFold loose on UniProt, a vast protein database used and updated by millions of researchers around the world. It has now predicted the structures of some 200 million proteins, almost all that are known to science.

Despite his success, Jumper remains modest about AlphaFold’s achievements. “That doesn’t mean that we’re certain of everything in there,” he says. “It’s a database of predictions, and it comes with all the caveats of predictions.”

A hard problem

Proteins are the biological machines that make living things work. They form muscles, horns, and feathers; they carry oxygen around the body and ferry messages between cells; they fire neurons, digest food, power the immune system; and so much more. But understanding exactly what a protein does (and what role it might play in various diseases or treatments) involves figuring out its structure—and that’s hard.

Proteins are made from strings of amino acids that chemical forces twist up into complex knots. An untwisted string gives few clues about the structure it will form. In theory, most proteins could take on an astronomical number of possible shapes. The task is to predict the correct one.

Jumper and his team built AlphaFold 2 using a type of neural network called a transformer, the same technology that underpins large language models. Transformers are very good at paying attention to specific parts of a larger puzzle.

But Jumper puts a lot of the success down to making a prototype model that they could test quickly. “We got a system that would give wrong answers at incredible speed,” he says. “That made it easy to start becoming very adventurous with the ideas you try.”

They stuffed the neural network with as much information about protein structures as they could, such as how proteins across certain species have evolved similar shapes. And it worked even better than they expected. “We were sure we had made a breakthrough,” says Jumper. “We were sure that this was an incredible advance in ideas.”

What he hadn’t foreseen was that researchers would download his software and start using it straight away for so many different things. Normally, it’s the thing a few iterations down the line that has the real impact, once the kinks have been ironed out, he says: “I’ve been shocked at how responsibly scientists have used it, in terms of interpreting it, and using it in practice about as much as it should be trusted in my view, neither too much nor too little.”

Any projects stand out in particular? 

Honeybee science

Jumper brings up a research group that uses AlphaFold to study disease resistance in honeybees. “They wanted to understand this particular protein as they look at things like colony collapse,” he says. “I never would have said, ‘You know, of course AlphaFold will be used for honeybee science.’”

He also highlights a few examples of what he calls off-label uses of AlphaFold—“in the sense that it wasn’t guaranteed to work”—where the ability to predict protein structures has opened up new research techniques. “The first is very obviously the advances in protein design,” he says. “David Baker and others have absolutely run with this technology.”

Baker, a computational biologist at the University of Washington, was a co-winner of last year’s chemistry Nobel, alongside Jumper and Hassabis, for his work on creating synthetic proteins to perform specific tasks—such as treating disease or breaking down plastics—better than natural proteins can.

Baker and his colleagues have developed their own tool based on AlphaFold, called RoseTTAFold. But they have also experimented with AlphaFold Multimer to predict which of their designs for potential synthetic proteins will work.    

“Basically, if AlphaFold confidently agrees with the structure you were trying to design then you make it and if AlphaFold says ‘I don’t know,’ you don’t make it. That alone was an enormous improvement.” It can make the design process 10 times faster, says Jumper.

Another off-label use that Jumper highlights: Turning AlphaFold into a kind of search engine. He mentions two separate research groups that were trying to understand exactly how human sperm cells hooked up with eggs during fertilization. They knew one of the proteins involved but not the other, he says: “And so they took a known egg protein and ran all 2,000 human sperm surface proteins, and they found one that AlphaFold was very sure stuck against the egg.” They were then able to confirm this in the lab.

“This notion that you can use AlphaFold to do something you couldn’t do before—you would never do 2,000 structures looking for one answer,” he says. “This kind of thing I think is really extraordinary.”

Five years on

When AlphaFold 2 came out, I asked a handful of early adopters what they made of it. Reviews were good, but the technology was too new to know for sure what long-term impact it might have. I caught up with one of those people to hear his thoughts five years on.

Kliment Verba is a molecular biologist who runs a lab at the University of California, San Francisco. “It’s an incredibly useful technology, there’s no question about it,” he tells me. “We use it every day, all the time.”

But it’s far from perfect. A lot of scientists use AlphaFold to study pathogens or to develop drugs. This involves looking at interactions between multiple proteins or between proteins and even smaller molecules in the body. But AlphaFold is known to be less accurate at making predictions about multiple proteins or their interaction over time.

Verba says he and his colleagues have been using AlphaFold long enough to get used to its limitations. “There are many cases where you get a prediction and you have to kind of scratch your head,” he says. “Is this real or is this not? It’s not entirely clear—it’s sort of borderline.”

“It’s sort of the same thing as ChatGPT,” he adds. “You know—it will bullshit you with the same confidence as it would give a true answer.”

Still, Verba’s team uses AlphaFold (both 2 and 3, because they have different strengths, he says) to run virtual versions of their experiments before running them in the lab. Using AlphaFold’s results, they can narrow down the focus of an experiment—or decide that it’s not worth doing.

It can really save time, he says: “It hasn’t really replaced any experiments, but it’s augmented them quite a bit.”

New wave  

AlphaFold was designed to be used for a range of purposes. Now multiple startups and university labs are building on its success to develop a new wave of tools more tailored to drug discovery. This year, a collaboration between MIT researchers and the AI drug company Recursion produced a model called Boltz-2, which predicts not only the structure of proteins but also how well potential drug molecules will bind to their target.  

Last month, the startup Genesis Molecular AI released another structure prediction model called Pearl, which the firm claims is more accurate than AlphaFold 3 for certain queries that are important for drug development. Pearl is interactive, so that drug developers can feed any additional data they may have to the model to guide its predictions.

AlphaFold was a major leap, but there’s more to do, says Evan Feinberg, Genesis Molecular AI’s CEO: “We’re still fundamentally innovating, just with a better starting point than before.”

Genesis Molecular AI is pushing margins of error down from less than two angstroms, the de facto industry standard set by AlphaFold, to less than one angstrom—one 10-millionth of a millimeter, or the width of a single hydrogen atom.

“Small errors can be catastrophic for predicting how well a drug will actually bind to its target,” says Michael LeVine, vice president of modeling and simulation at the firm. That’s because chemical forces that interact at one angstrom can stop doing so at two. “It can go from ‘They will never interact’ to ‘They will,’” he says.

With so much activity in this space, how soon should we expect new types of drugs to hit the market? Jumper is pragmatic. Protein structure prediction is just one step of many, he says: “This was not the only problem in biology. It’s not like we were one protein structure away from curing any diseases.”

Think of it this way, he says. Finding a protein’s structure might previously have cost $100,000 in the lab: “If we were only a hundred thousand dollars away from doing a thing, it would already be done.”

At the same time, researchers are looking for ways to do as much as they can with this technology, says Jumper: “We’re trying to figure out how to make structure prediction an even bigger part of the problem, because we have a nice big hammer to hit it with.”

In other words, make everything into nails? “Yeah, let’s make things into nails,” he says. “How do we make this thing that we made a million times faster a bigger part of our process?”

What’s next?

Jumper’s next act? He wants to fuse the deep but narrow power of AlphaFold with the broad sweep of LLMs.  

“We have machines that can read science. They can do some scientific reasoning,” he says. “And we can build amazing, superhuman systems for protein structure prediction. How do you get these two technologies to work together?”

That makes me think of a system called AlphaEvolve, which is being built by another team at Google DeepMind. AlphaEvolve uses an LLM to generate possible solutions to a problem and a second model to check them, filtering out the trash. Researchers have already used AlphaEvolve to make a handful of practical discoveries in math and computer science.    

Is that what Jumper has in mind? “I won’t say too much on methods, but I’ll be shocked if we don’t see more and more LLM impact on science,” he says. “I think that’s the exciting open question that I’ll say almost nothing about. This is all speculation, of course.”

Jumper was 39 when he won his Nobel Prize. What’s next for him?

“It worries me,” he says. “I believe I’m the youngest chemistry laureate in 75 years.” 

He adds: “I’m at the midpoint of my career, roughly. I guess my approach to this is to try to do smaller things, little ideas that you keep pulling on. The next thing I announce doesn’t have to be, you know, my second shot at a Nobel. I think that’s the trap.”

MIT Technology Review’s What’s Next series looks across industries, trends, and technologies to give you a first look at the future. You can read the rest of them here.

In the early 2020s, a little-known aquaculture company in Portland, Maine, snagged more than $50 million by pitching a plan to harness nature to fight back against climate change. The company, Running Tide, said it could sink enough kelp to the seafloor to sequester a billion tons of carbon dioxide by this year, according to one of its early customers.

Instead, the business shut down its operations last summer, marking the biggest bust to date in the nascent carbon removal sector.

Its demise was the most obvious sign of growing troubles and dimming expectations for a space that has spawned hundreds of startups over the last few years. A handful of other companies have shuttered, downsized, or pivoted in recent months as well. Venture investments have flagged. And the collective industry hasn’t made a whole lot more progress toward that billion-ton benchmark.

The hype phase is over and the sector is sliding into the turbulent business trough that follows, warns Robert Höglund, cofounder of CDR.fyi, a public-benefit corporation that provides data and analysis on the carbon removal industry.

“We’re past the peak of expectations,” he says. “And with that, we could see a lot of companies go out of business, which is natural for any industry.”

The open question is: If the carbon removal sector is heading into a painful if inevitable clearing-out cycle, where will it go from there? 

The odd quirk of carbon removal is that it never made a lot of sense as a business proposition: It’s an atmospheric cleanup job, necessary for the collective societal good of curbing climate change. But it doesn’t produce a service or product that any individual or organization strictly needs—or is especially eager to pay for.

To date, a number of businesses have voluntarily agreed to buy tons of carbon dioxide that companies intend to eventually suck out of the air. But whether they’re motivated by sincere climate concerns or pressures from investors, employees, or customers, corporate do-goodism will only scale any industry so far. 

Most observers argue that whether carbon removal continues to bobble along or transforms into something big enough to make a dent in climate change will depend largely on whether governments around the world decide to pay for a whole, whole lot of it—or force polluters to. 

“Private-sector purchases will never get us there,” says Erin Burns, executive director of Carbon180, a nonprofit that advocates for the removal and reuse of carbon dioxide. “We need policy; it has to be policy.”

What’s the problem?

The carbon removal sector began to scale up in the early part of this decade, as increasingly grave climate studies revealed the need to dramatically cut emissions and suck down vast amounts of carbon dioxide to keep global warming in check.

Specifically, nations may have to continually remove as much as 11 billion tons of carbon dioxide per year by around midcentury to have a solid chance of keeping the planet from warming past 2 °C over preindustrial levels, according to a UN climate panel report in 2022.

A number of startups sprang up to begin developing the technology and building the infrastructure that would be needed, trying out a variety of approaches like sinking seaweed or building carbon-dioxide-sucking factories.

And they soon attracted customers. Companies including Stripe, Google, Shopify, Microsoft, and others began agreeing to pre-purchase tons of carbon removal, hoping to stand up the nascent industry and help offset their own climate emissions. Venture investments also flooded into the space, peaking in 2023 at nearly $1 billion, according to data provided by PitchBook.

From early on, players in the emerging sector sought to draw a sharp distinction between conventional carbon offset projects, which studies have shown frequently exaggerate climate benefits, and “durable” carbon removal that could be relied upon to suck down and store away the greenhouse gas for decades to centuries. There’s certainly a big difference in the price: While buying carbon offsets through projects that promise to preserve forests or plant trees might cost a few dollars per ton, a ton of carbon removal can run hundreds to thousands of dollars, depending on the approach. 

That high price, however, brings big challenges. Removing 10 billion tons of carbon dioxide a year at, say, $300 a ton adds up to a global price tag of $3 trillion—a year. 

Which brings us back to the fundamental question: Who should or would foot the bill to develop and operate all the factories, pipelines, and wells needed to capture, move, and bury billions upon billions of tons of carbon dioxide?

The state of the market

The market is still growing, as companies voluntarily purchase tons of carbon removal to make strides toward their climate goals. In fact, sales reached an all-time high in the second quarter of this year, mostly thanks to several massive purchases by Microsoft.

But industry sources fear that demand isn’t growing fast enough to support a significant share of the startups that have formed or even the projects being built, undermining the momentum required to scale the sector up to the size needed by midcentury.

To date, all those hundreds of companies that have spun up in recent years have disclosed deals to sell some 38 million tons of carbon dioxide pulled from the air, according to CDR.fyi. That’s roughly the amount the US pumps out in energy-related emissions every three days. 

And they’ve only delivered around 940,000 tons of carbon removal. The US emits that much carbon dioxide in less than two hours. (Not every transaction is publicly announced or revealed to CDR.fyi, so the actual figures could run a bit higher.)

Another concern is that the same handful of big players continue to account for the vast majority of the overall purchases, leaving the health and direction of the market dependent on their whims and fortunes. 

Most glaringly, Microsoft has agreed to buy 80% of all the carbon removal purchased to date, according to  CDR.fyi. The second-biggest buyer is Frontier, a coalition of companies that includes Google, Meta, Stripe, and Shopify, which has committed to spend $1 billion.

If you strip out those two buyers, the market shrinks from 16 million tons under contract during the first half of this year to just 1.2 million, according to data provided to MIT Technology Review by CDR.fyi. 

Signs of trouble

Meanwhile, the investor appetite for carbon removal is cooling. For the 12-month period ending in the second quarter of 2025, venture capital investments in the sector fell more than 13% from the same period last year, according to data provided by PitchBook. That tightening funding will make it harder and harder for companies that aren’t bringing in revenue to stay afloat.

Other companies that have already shut down include the carbon removal marketplace Nori, the direct air capture company Noya and Alkali Earth, which was attempting to use industrial by-products to tie up carbon dioxide.

Still other businesses are struggling. Climeworks, one of the first companies to build direct-air-capture (DAC) factories, announced it was laying off 10% of its staff in May, as it grapples with challenges on several fronts.

The company’s plans to collaborate on the development of a major facility in the US have been at least delayed as the Trump administration has held back tens of millions of dollars in funding granted in 2023 under the Department of Energy’s Regional Direct Air Capture Hubs program. It now appears the government could terminate the funding altogether, along with perhaps tens of billions of dollars’ worth of additional grants previously awarded for a variety of other US carbon removal and climate tech projects.

“Market rumors have surfaced, and Climeworks is prepared for all scenarios,” Christoph Gebald, one of the company’s co-CEOs, said in a previous statement to MIT Technology Review. “The need for DAC is growing as the world falls short of its climate goals and we’re working to achieve the gigaton capacity that will be needed.”

But purchases from direct-air-capture projects fell nearly 16% last year and account for just 8% of all carbon removal transactions to date. Buyers are increasingly looking to categories that promise to deliver tons faster and for less money, notably including burying biochar or installing carbon capture equipment on bioenergy plants. (Read more in my recent story on that method of carbon removal, known as BECCS, here.)

CDR.fyi recently described the climate for direct air capture in grim terms: “The sector has grown rapidly, but the honeymoon is over: Investment and sales are falling, while deployments are delayed across almost every company.”

“Most DAC companies,” the organization added, “will fold or be acquired.”

What’s next?

In the end, most observers believe carbon removal isn’t really going to take off unless governments bring their resources and regulations to bear. That could mean making direct purchases, subsidizing these sectors, or getting polluters to pay the costs to do so—for instance, by folding carbon removal into market-based emissions reductions mechanisms like cap-and-trade systems. 

More government support does appear to be on the way. Notably, the European Commission recently proposed allowing “domestic carbon removal” within its EU Emissions Trading System after 2030, integrating the sector into one of the largest cap-and-trade programs. The system forces power plants and other polluters in member countries to increasingly cut their emissions or pay for them over time, as the cap on pollution tightens and the price on carbon rises. 

That could create incentives for more European companies to pay direct-air-capture or bioenergy facilities to draw down carbon dioxide as a means of helping them meet their climate obligations.

There are also indications that the International Civil Aviation Organization, a UN organization that establishes standards for the aviation industry, is considering incorporating carbon removal into its market-based mechanism for reducing the sector’s emissions. That might take several forms, including allowing airlines to purchase carbon removal to offset their use of traditional jet fuel or requiring the use of carbon dioxide obtained through direct air capture in some share of sustainable aviation fuels.

Meanwhile, Canada has committed to spend $10 million on carbon removal and is developing a protocol to allow direct air capture in its national offsets program. And Japan will begin accepting several categories of carbon removal in its emissions trading system. 

Despite the Trump administration’s efforts to claw back funding for the development of carbon-sucking projects, the US does continue to subsidize storage of carbon dioxide, whether it comes from power plants, ethanol refineries, direct-air-capture plants, or other facilities. The so-called 45Q tax credit, which is worth up to $180 a ton, was among the few forms of government support for climate-tech-related sectors that survived in the 2025 budget reconciliation bill. In fact, the subsidies for putting carbon dioxide to other uses increased.

Even in the current US political climate, Burns is hopeful that local or federal legislators will continue to enact policies that support specific categories of carbon removal in the regions where they make the most sense, because the projects can provide economic growth and jobs as well as climate benefits.

“I actually think there are lots of models for what carbon removal policy can look like that aren’t just things like tax incentives,” she says. “And I think that this particular political moment gives us the opportunity in a unique way to start to look at what those regionally specific and pathway specific policies look like.”

The dangers ahead

But even if more nations do provide the money or enact the laws necessary to drive the business of durable carbon renewal forward, there are mounting concerns that a sector conceived as an alternative to dubious offset markets could increasingly come to replicate their problems.

Various incentives are pulling in that direction.

Financial pressures are building on suppliers to deliver tons of carbon removal. Corporate buyers are looking for the fastest and most affordable way of hitting their climate goals. And the organizations that set standards and accredit carbon removal projects often earn more money as the volume of purchases rises, creating clear conflicts of interest.

Some of the same carbon registries that have long signed off on carbon offset projects have begun creating standards or issuing credits for various forms of carbon removal, including Verra and Gold Standard.

“Reliable assurance that a project’s declared ton of carbon savings equates to a real ton of emissions removed, reduced, or avoided is crucial,” Cynthia Giles, a senior EPA advisor under President Biden, and Cary Coglianese, a law professor at the University of Pennsylvania, wrote in a recent editorial in Science. “Yet extensive research from many contexts shows that auditors selected and paid by audited organizations often produce results skewed toward those entities’ interests.”

Noah McQueen, the director of science and innovation at Carbon180, has stressed that the industry must strive to counter the mounting credibility risks, noting in a recent LinkedIn post: “Growth matters, but growth without integrity isn’t growth at all.”

In an interview, McQueen said that heading off the problem will require developing and enforcing standards to truly ensure that carbon removal projects deliver the climate benefits promised. McQueen added that to gain trust, the industry needs to earn buy-in from the communities in which these projects are built and avoid the environmental and health impacts that power plants and heavy industry have historically inflicted on disadvantaged communities.

Getting it right will require governments to take a larger role in the sector than just subsidizing it, argues David Ho, a professor at the University of Hawaiʻi at Mānoa who focuses  on ocean-based carbon removal.

He says there should be a massive, multinational research drive to determine the most effective ways of mopping up the atmosphere with minimal environmental or social harm, likening it to a Manhattan Project (minus the whole nuclear bomb bit).

“If we’re serious about doing this, then let’s make it a government effort,” he says, “so that you can try out all the things, determine what works and what doesn’t, and you don’t have to please your VCs or concentrate on developing [intellectual property] so you can sell yourself to a fossil-fuel company.”

Ho adds that there’s a moral imperative for the world’s historically biggest climate polluters to build and pay for the carbon-sucking and storage infrastructure required to draw down billions of tons of greenhouse gas. That’s because the world’s poorest, hottest nations, which have contributed the least to climate change, will nevertheless face the greatest dangers from intensifying heat waves, droughts, famines, and sea-level rise.

“It should be seen as waste management for the waste we’re going to dump on the Global South,” he says, “because they’re the people who will suffer the most from climate change.”

Correction (October 24): An earlier version of this article referred to Noya as a carbon removal marketplace. It was a direct air capture company.