AI Mornings with Andreas Vig
Your daily AI news briefing in under 10 minutes. New models, product launches, research breakthroughs, and industry shifts, explained clearly, no hype.
AI Mornings with Andreas Vig
Cerebras' $20B OpenAI Deal & AI Fraud Charges
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Hey, welcome to AI Mornings with Andreas Vig. It's April 20th, 2026. The biggest story today comes from the AI hardware space. Cerebrus Systems, the chip startup known for building processors the size of dinner plates, has filed to go public. This comes after the company secured what might be the largest AI infrastructure deal ever announced. OpenAI has committed to pay Cerebrus more than 20 billion US dollars over the next three years to use their servers, deploying 750 megawatts of computing power through 2028. OpenAI also loaned the company$1 billion and received warrants to buy stock. Cerebrus CEO Andrew Feldman didn't mince words about the competitive implications, saying they took the fast inference business from Nvidia at OpenAI. The company has been on a fundraising tear, raising a$1.1 billion Series G and then a$1 billion Series H at a$23 billion valuation. They reported$510 million in revenue for 2025. The IPO is expected in mid-May. Also in legal news, federal prosecutors have charged the former CEO and CFO of iLearning Engines with running what they're calling a massive fraud scheme. The indictment alleges that the executives fabricated virtually all of the company's customer relationships and revenue, with at least 90% of their reported$421 million in 2023 revenue being completely fake. The company went public in April 2024 and peaked at a$1.5 billion market valuation before short seller Hindenburg Research exposed the fraud. Prosecutors say the executives used forged contracts in what they call round-trip money transfers, where they sent money to fake customers who then returned it to create the appearance of revenue. The company is now bankrupt, and both executives face a 1-0 count indictment, including securities fraud, wire fraud, and conspiracy. In policy news, Anthropic appears to be patching things up with the Trump administration. CEO Dario Amode met with Treasury Secretary Scott Bissent and White House Chief of Staff Susie Wiles in what the White House described as a productive and constructive meeting. This comes after the Pentagon designated Anthropic as a supply chain risk, a label usually reserved for foreign adversaries, because the company refused to allow unrestricted military use of its models for autonomous weapons and mass surveillance. An administration source told Axios that every agency except the Department of Defense wants to use Anthropic's technology. The company is still challenging the Pentagon designation in court, but it seems the rest of the administration is eager to work with them. Moving to autonomous vehicles, Uber is making a massive bet on Robotaxis. The company has committed over$10 billion to partners including Baidu, Rivian, and Lucid to build out driverless services. The Financial Times reports this is a sharp pivot from the gig economy model that built Uber's empire, driven by fears that robotaxi rivals could disrupt their business. They're targeting 28 cities by 2028. Meanwhile, Tesla has expanded its own robotaxi service to Dallas and Houston, bringing its driverless operations to three Texas cities. Though the rollout appears limited, crowdsourced data shows only one vehicle active in each new city compared to 46 inches Austin. Tesla's Austin Robotaxis have been involved in 14 crashes since launch. Alright, a few more things worth knowing about today. In robotics, Unitree has released footage of its H1 humanoid robot sprinting at 10.1 meters per second, which puts it within striking distance of Usain Bolt's world record pace. The company thinks humanoid robots could break the 1-0, second 100 meter barrier by mid-2026. Just two years ago, the fastest humanoid could barely manage a brisk walk. And Toyota's Q7 basketball robot is making waves for sinking near-perfect shots using reinforcement learning, serving as a testbed for embodied AI and vision systems. On the research front, scientists have built a computer that runs entirely without electricity, powered instead by the physical memory of springs and steel. The mechanical devices can count and retain measurements without any power source, which could be transformative for environments that destroy traditional hardware. Separately, researchers at Northwestern University have printed flexible artificial neurons that can fire signals real brain cells actually respond to, triggering responses in live mouse brain tissue. Finally, a couple of industry notes. A new PWC study found that 75% of AI's economic gains are being captured by just 20% of companies, with leaders focused on growth rather than just productivity. And the European Union has finalized AI hiring bias audit rules that take effect in 105 days, covering any AI tool used to evaluate EU based candidates regardless of where the company is headquartered. That's it for today. See you tomorrow.