Daily Signal: April 29, 2026
Algorithm Times' Daily Signal is a daily sweep of the AI headlines worth reading, with context for why they matter.
Today, the broad theme around the renegotiation of AI's foundational contracts continues, and the first hard earnings data from Big Tech AI earnings week are in.
There's also a capital rotation toward reinforcement learning that's now large enough to track as a structural trend.
Microsoft's Azure growth reaccelerated to 40% in Q3 and its AI business crossed a $37 billion annualized run rate
Microsoft posted Q3 fiscal 2026 revenue of $82.9 billion, up 18% year over year. Azure grew 40%, above the company's own 37-38% guidance range, and the total AI business reached a $37 billion annualized run rate, up 123% year over year.
Commercial remaining performance obligation hit $627 billion, up 99%. These are the first verified data points of Big Tech AI earnings week; Alphabet, Amazon, and Meta report tonight. The Azure beat matters because it suggests enterprise AI workloads are driving cloud acceleration rather than substituting for existing spend, and the 99% RPO growth indicates customers are locking in multi-year commitments at a rate that looks structural rather than cyclical.
The $37 billion AI run rate also establishes a concrete benchmark against which OpenAI's own revenue trajectory, reported as underperforming this week, will be measured.
OpenAI and Microsoft restructured into a non-exclusive arrangement, and OpenAI models went live on Amazon Bedrock days later
Under the revised partnership terms, Microsoft holds a non-exclusive license to OpenAI IP through 2032, Microsoft stops paying OpenAI a revenue share while OpenAI continues paying Microsoft through 2030 subject to an undisclosed cap, and OpenAI must still ship new models to Azure first before other clouds.
OpenAI also committed to approximately 2 gigawatts of AWS Trainium capacity under the Amazon agreement. One day after the restructuring closed, OpenAI confirmed its models, including the Codex coding agent, are now accessible through Amazon Bedrock in limited preview, with general availability expected within weeks, alongside a co-developed Amazon Bedrock Managed Agents service.
For enterprise engineering teams currently running on Azure, the Codex agent is now accessible without an Azure dependency. The first-ship-to-Azure clause is the key constraint to monitor as new model releases approach, since it determines how long any competitive advantage from the Bedrock launch actually holds.
OpenAI missed multiple monthly revenue targets, and its one billion user goal, and its CFO flagged near-term risk on data center funding
OpenAI missed several monthly sales targets in early 2026 and did not reach its internal goal of one billion weekly active ChatGPT users by the end of 2025, per the Wall Street Journal and Bloomberg reporting.
CFO Sarah Friar privately warned colleagues the company may struggle to fund its data center commitments if revenue growth does not accelerate. The infrastructure commitments are not speculative: OpenAI holds a $250 billion Azure commitment through 2032 and a $300 billion five-year deal with Oracle, with projected operating losses of roughly $74 billion in 2028 before an anticipated return to profit by 2030.
Anthropic gained ground in coding and enterprise markets during the same period. The CFO warning introduces a specific near-term funding risk, distinct from the long-dated structural one that has always been part of the OpenAI story, and it arrives in the same week the company added new infrastructure commitments via the AWS agreement rather than reducing them. Compute partners, including Oracle and CoreWeave, are downstream of this revenue trajectory.
On day three of the Oakland trial, Elon Musk testified that the Microsoft investment was the moment OpenAI abandoned its nonprofit mission
Musk underwent contentious cross-examination from OpenAI's attorneys and testified that Microsoft's $10 billion investment was the tipping point, convincing him OpenAI had violated its nonprofit charter.
The jury's verdict will advise Judge Yvonne Gonzalez Rogers on whether to award $130 billion in damages, revert OpenAI to nonprofit status, and remove Sam Altman and Greg Brockman from the board. Any of those outcomes directly affects the planned IPO timeline.
The trial is running concurrently with the revenue shortfall reporting and the multi-cloud restructuring, which means the governance, business, and infrastructure risks are compressing into the same window rather than arriving sequentially.
What would otherwise be a background legal proceeding is now a live structural threat to a company renegotiating its core partnerships under financial pressure.
Google extended its Pentagon contract to cover classified networks the day after 600-plus employees, including DeepMind researchers, signed a letter objecting
Google amended its existing DoD contract to extend Gemini's availability to classified networks under terms permitting "any lawful government purpose," one day after more than 600 employees urged CEO Sundar Pichai to refuse.
Bloomberg separately reported that Google withdrew from a $100 million Pentagon drone swarm prize challenge in February following an internal ethics review, a detail that complicates any reading of this week's move as a clean policy shift.
The amended contract places Google alongside OpenAI and xAI on classified DoD networks. Anthropic declined identical terms. For federal AI procurement teams, the competitive landscape now has a functional division: three major labs operating inside the classified perimeter, and Anthropic outside it, at least for now.
The White House is drafting guidance to restore Anthropic to federal procurement without requiring the classified-use concession that triggered the standoff
Axios and Nextgov report that the White House is developing guidance that would allow agencies to work around Anthropic's supply chain risk designation and onboard new models, including Mythos.
A draft executive action is under discussion following a meeting between Anthropic CEO Dario Amodei and Chief of Staff Susie Wiles; sources describe the intent as a way to "save face and bring them back in."
This lands on the same day Google signed the Pentagon deal Anthropic declined, compressing the policy window considerably. If the executive action advances, it reopens the federal market to Anthropic's models without the classified-use language that caused the break, materially changing the procurement calculus for agencies that have been routing around Anthropic for the past several months.
David Silver's Ineffable Intelligence launched with a $1.1 billion seed at a $5.1 billion valuation, the largest seed round in European AI history
Former DeepMind reinforcement learning lead David Silver launched UK-based Ineffable Intelligence with a round co-led by Sequoia and Lightspeed, with Nvidia, Google, and the UK Sovereign AI Fund on the cap table. The company is pursuing superintelligence through reinforcement learning rather than by scaling LLMs.
This is the third major RL-focused raise in roughly 90 days: AMI Labs, co-founded by Yann LeCun, raised $1.03 billion at a $3.5 billion valuation last month, and Recursive Superintelligence, co-founded by former DeepMind principal scientist Tim Rocktaschel, reportedly raised approximately $500 million with demand sufficient to stretch toward $1 billion.
Three UK-linked labs, founded by verifiable senior researchers with published RL or post-LLM theses, raised a combined $2.6 billion in seed funding in roughly three months. The presence of Nvidia and Google on the Ineffable cap table suggests a strategic positioning against the risk of an LLM scaling plateau, not just financial exposure to upside.
