Brain cells learn Doom gameplay & OpenAI lands massive funding round - News (Mar 1, 2026)

Let’s start with the day’s most mind-bending tech story. An Australian company, Cortical Labs, says it has trained living human neurons—grown directly on a microchip—to play the classic first-person shooter Doom. The setup uses microelectrode arrays to both stimulate the neurons and read their electrical activity, turning a dish of living cells into a kind of “biological computer.” What’s new this time isn’t just the game choice—Doom is dramatically more complex than Pong, which the same group showcased back in 2021. The bigger shift is accessibility: Cortical Labs built a new interface that developers can program using Python. An independent developer with limited biology background reportedly used those tools to get the neuron-chip interacting with Doom within days. Researchers say performance isn’t anywhere near human-level gaming, but it beat random behavior, and some experts see the demo as a real step forward in training living neural systems. The unanswered question is still a big one: we can observe the outputs, but we don’t fully understand how the neurons are representing the task—especially something as visual as a game screen.

Now to the major headline in AI business: OpenAI says it has secured $110 billion in new funding, valuing the company at a $730 billion pre-money valuation. According to CEO Sam Altman, Amazon is leading the round with a $50 billion commitment, while Nvidia and SoftBank have each pledged $30 billion. The Amazon money is structured: an initial $15 billion, followed by another $35 billion over the coming months if preset conditions are met. Altman also says additional investors are expected to join as the round continues. He framed the cash and partnerships as fuel for global expansion, bigger infrastructure, and a stronger balance sheet—essentially, the plumbing required to bring what he calls “frontier AI” into everyday use at worldwide scale. Altman also shared new usage numbers: ChatGPT now reportedly has more than 900 million weekly active users and over 50 million consumer subscribers. Whether you see that as exhilarating, unsettling, or both, it underlines a shift the industry keeps hinting at: the race isn’t only about smarter models—it’s about who can scale reliably and turn that scale into products people actually depend on.

That scaling story is tightly tied to cloud and compute—and OpenAI is reshaping its partnerships. As part of a multiyear deal with Amazon, AWS will become the exclusive third-party cloud distribution provider for OpenAI Frontier. The companies say they’ll jointly deliver advanced AI capabilities for enterprises. OpenAI and AWS are also expanding an existing multiyear agreement—reportedly adding $100 billion over eight years—plus work on customized models for Amazon developers building customer-facing applications. OpenAI also says it’s expanding its partnership with Nvidia. At the same time, the company insists its long-standing relationship with Microsoft remains “strong and central.” In other words: OpenAI is widening the circle, but it’s not publicly cutting old ties.

Stepping back, a broader infrastructure boom is accelerating alongside AI products. TechCrunch reports that the real frenzy now includes data centers, power generation, GPUs, and the construction capacity to keep it all moving. Nvidia CEO Jensen Huang has projected that $3 to $4 trillion could be spent on AI infrastructure by the end of the decade. The piece traces how today’s compute alignments were shaped by early deals—like Microsoft’s first big OpenAI investment and the way cloud credits became a form of financing as training costs exploded. It also points to increasingly unconventional arrangements: GPU-for-equity structures, giant multi-year compute contracts, and a kind of circular market where scarce chips and scarce private stock are exchanged to keep expansion on schedule. There’s also a harder edge to this story: environmental impact and local backlash. When you mix data-center growth with power plants, emissions, and grid constraints, AI stops being a purely digital conversation. The big question hanging over all of it is straightforward: can the industry prove that this extraordinary spending will generate returns that justify the scale—and the tradeoffs?

On the government side of AI, OpenAI says it has now reached an agreement with the U.S. Department of War—the Pentagon—after a week of very public tensions involving rival Anthropic. According to reporting by Fortune, Altman told employees the government will allow OpenAI to keep control of its “safety stack,” and would not force the company to override model refusals. Altman also described key limits: OpenAI would choose which models are deployed and where, and deployments would be restricted to cloud environments—not “edge systems” like aircraft or drones. Most notably, he said the Pentagon would include OpenAI’s “red lines” in the contract, including prohibitions on using AI for autonomous weapons, domestic mass surveillance, or critical decision-making. This comes as President Trump ordered federal agencies to stop using Anthropic technology, with a six-month phase-out. The bigger takeaway is that Washington is trying to standardize what it can demand from AI providers—and AI providers are trying to define what they simply won’t do, even under government pressure.

Turning to health and medicine, European drug regulators have backed a major simplification in treating sleeping sickness—also known as human African trypanosomiasis. A committee at the European Medicines Agency endorsed acoziborole, a pill made by Sanofi, in a step expected to help the drug reach patients first in the Democratic Republic of Congo and later across other affected regions. The potential breakthrough is practical: it’s taken as three pills at once, in a single dose. That’s a dramatic change from current approaches that can require long hospital stays, complex staging, and even spinal taps to determine the right regimen. Sleeping sickness is spread by tsetse flies in rural sub-Saharan Africa. Early symptoms can mimic the flu, but untreated infections can invade the nervous system, disrupt sleep cycles, and become fatal. A vaccine has been difficult because the parasite can change its outer proteins, dodging immune responses. In a study of about 200 patients in Congo and Guinea, more than 95% of those treated with acoziborole were considered cured 18 months later. The drug can be used for people 12 and older, and for both early and advanced infections—removing the need for spinal taps. Sanofi says it will donate doses to the World Health Organization so patients receive it for free. Researchers caution there are still unknowns, including where the parasite may be hiding, but supporters argue this could meaningfully speed efforts to stop transmission by 2030.

Another medical development could reshape cancer screening down the road. A Danish research team studied why the gut bacterium Bacteroides fragilis is often associated with colorectal cancer even though it’s common in healthy people. Their finding: cancer-linked strains didn’t appear to come from a single inherited bacterial lineage. Instead, the strongest signal involved “prophages”—viruses that embed themselves into bacterial DNA and can change how bacteria behave. The team identified two previously unknown prophages that were common in B. fragilis from colorectal cancer patients, but mostly absent in people without cancer. In a broader analysis of fecal microbiome data from 877 people across multiple regions, cancer patients were more than twice as likely to have detectable levels of these prophage signatures. This is correlation, not proof that the prophages cause cancer—and the researchers openly note alternative explanations, like cancer changing the gut environment. But it raises a compelling screening idea: adding viral markers to stool tests. In an early analysis, a prophage-fragment panel detected about 40% of cases—nowhere near good enough on its own, but potentially useful as part of a combined approach if validated further.

Now to geopolitics, where tensions around Iran continue to escalate—and the details are moving fast. The Japan Times reports that the U.S. military used kamikaze-style “one-way attack” drones in strikes on Iran over the weekend, marking the first time the United States has used this category of weapon in combat. U.S. Central Command said Task Force Scorpion Strike employed the drones during “Operation Epic Fury,” described as part of a joint operation with Israel. Central Command framed the drones as low-cost systems modeled after Iran’s Shahed designs—calling it, quote, “American-made retribution.” An image released by the U.S. showed the Low-cost Unmanned Combat Attack System, or LUCAS, which the reporting notes is derived from the Shahed-136 concept that has also been used widely in Ukraine by Russian forces. Meanwhile, Canada’s Prime Minister Mark Carney says Ottawa supports the U.S. objective of destroying Iran’s nuclear program, while emphasizing Canada is not participating militarily and was not involved in planning. Carney called Iran a major source of regional instability, but Canadian politics are split: Conservatives voiced support for Iranians overthrowing the regime, while the NDP condemned the bombings as a dangerous escalation, and the Bloc raised concerns about legality and the lack of U.S. congressional approval. And hovering over all of this is the nuclear verification problem. A confidential IAEA report says Iran has not allowed inspectors access to nuclear facilities bombed by Israel and the U.S. during a 12-day war in June. Without access, the agency says it cannot verify whether enrichment-related work has stopped or confirm the size and location of uranium stockpiles at affected sites. The IAEA estimates Iran has about 440.9 kilograms enriched up to 60% purity—close to weapons grade—an amount the agency says could be enough for as many as 10 bombs if weaponized, though it stresses that doesn’t mean a weapon exists. The IAEA has been relying on satellite imagery and notes activity around sites like Isfahan, Natanz, and Fordow, but says it cannot confirm what that activity means. Negotiations are still underway, with technical discussions expected to continue next week in Vienna. The bottom line: without inspections, everyone is arguing from partial information—and that makes miscalculation more likely.

Finally, U.S. trade policy is back in the spotlight after a major legal blow and an immediate workaround. The U.S. Supreme Court struck down President Trump’s broad tariff measures, ruling the International Emergency Economic Powers Act does not authorize sweeping import tariffs. But the uncertainty didn’t end there. Within a day, Trump signaled he would keep pressing ahead and invoked a temporary global tariff—starting at 10% and then raised to 15%, described as the maximum permitted under U.S. trade law. The European Commission rejected any tariff hike and insisted existing agreements must be respected. India postponed a planned trade visit to Washington to reassess what the shift means. Markets treated the back-and-forth as destabilizing. The U.S. dollar moved unevenly, and investors leaned into traditional hedges: gold futures surged above $5,200, and silver jumped sharply. Energy and industrial metals also face cross-currents, since currency swings and trade disruption can reshape demand forecasts and supply chains. One notable angle in the reporting: the more unpredictable tariff policy becomes, the more buyers like India may lean toward discounted Russian crude for price stability.