AI News Today — July 23, 2026:
The 5 Biggest Stories You Need to Know
AMD just wrote a $5 billion check to Anthropic and committed 2 gigawatts of its next-generation MI450 GPUs — the largest chip-for-equity deal in AI history and a direct challenge to Nvidia’s stranglehold on frontier training compute. The same day, Alphabet reported Q2 earnings showing Google Cloud revenue up 82% and Gemini crossing 950 million monthly users. Meanwhile, the White House publicly accused Moonshot AI of distilling Anthropic’s Fable model to build Kimi K3, monday.com cut 20% of its workforce to reorg around AI, and the Trump administration committed $5 billion to a new AI-for-science initiative. Here’s everything that moved.
AMD Backs Anthropic With $5 Billion and 2 Gigawatts of MI450 Chips
AMD and Anthropic announced a strategic partnership at AMD’s Advancing AI event on July 22 to deploy up to 2 gigawatts of AMD Instinct MI450 Series GPUs in AMD Helios rack-scale systems, with the first gigawatt going online in H1 2027. AMD will make a strategic equity investment of up to $5 billion in Anthropic. The systems pair MI455X GPUs with the newly launched 256-core EPYC Venice CPUs (Zen 6, TSMC 2nm) and Pensando networking. Anthropic will use Claude to help accelerate AMD’s ROCm software development.
This is the deal that reshapes the compute landscape. For two years, every frontier lab has been locked into Nvidia’s ecosystem — H100s, then B200s, then GB300s — with pricing power that let Nvidia capture most of the value in the AI stack. AMD just offered Anthropic a way out: $5 billion in equity plus 2 gigawatts of next-gen silicon. Lisa Su called it “the full strength of AMD high-performance computing.” If Anthropic can train Claude on MI450s at competitive performance, every other lab will have leverage against Nvidia for the first time. AMD also debuted EPYC Venice — 256 cores, TSMC 2nm, Gate-All-Around transistors — claiming a 70% performance leap over the previous generation. Systems ship Q3 2026.
Alphabet Q2: Google Cloud Up 82%, Gemini Crosses 950 Million Users
Alphabet reported Q2 revenue of $119.8 billion, up 24% year-over-year. Google Cloud accelerated to $24.77 billion in revenue — an 82% jump — with cloud backlog reaching $514 billion, up from $462 billion in Q1. Gemini App hit 950 million monthly active users, up from 900 million in May and 750 million at the end of 2025. Gemini models now process 22 billion API tokens per minute. Capital expenditure doubled to $44.9 billion for the quarter. YouTube ad revenue rose 13% to $11.06 billion.
The cloud number is staggering — 82% growth at that scale means Google Cloud is now a $100 billion annualized run rate business. The Gemini user count is equally significant: 950 million monthly actives means Gemini is the most widely used AI assistant in the world by a wide margin, even if engagement depth per user likely trails ChatGPT. The $44.9 billion in quarterly capex — nearly $180 billion annualized — shows Google is spending more on AI infrastructure in a single quarter than most companies are worth. All of this lands while Gemini 3.5 Pro still hasn’t shipped, which makes the usage growth even more notable.
White House Officially Accuses Moonshot AI of Distilling Anthropic’s Fable to Build K3
White House OSTP Director Michael Kratsios posted on X on July 22 that “we have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model,” calling it “covert industrial distillation aimed at stealing proprietary U.S. technology.” This is the first named public accusation from a senior US official that ties Kimi K3 specifically to Fable. Separately, OpenAI President Greg Brockman told Bloomberg he considers K3 “pretty good” but said it’s “too early” to conclude whether distillation occurred.
Kratsios’s accusation also claims Moonshot obtained Nvidia GB300 chips via Thailand to train K3, sidestepping US export controls. No supporting evidence was disclosed publicly. The geopolitical significance is clear: the US government is now willing to publicly name a Chinese AI lab and accuse it of IP theft from a specific American company (Anthropic). Brockman’s more measured response — calling K3 “pretty good” without making accusations — is the interesting contrast. If the distillation claims are substantiated, this escalates the AI trade war dramatically. If they aren’t, it sets a precedent for using unsubstantiated IP claims as trade policy.
monday.com Cuts 20% of Workforce to Reorganize Entirely Around AI
monday.com announced on July 22 that it will lay off approximately 620 employees — 20% of its global workforce, including 350 in Tel Aviv — as it flattens management layers and reorganizes around what co-CEOs Roy Mann and Eran Zinman called “the largest product change in the company’s history.” The redesign centers the platform around human-and-agent collaboration. MNDY stock is down roughly 50% year-to-date and more than 80% from its all-time peak.
The company insists these are not cost cuts and not human-for-AI replacement, but a resource reallocation toward AI product development. That’s a distinction that matters to executives writing press releases and not at all to the 620 people who lost their jobs. This follows Oracle’s 30,000-person cut last week and Amazon’s layoffs in its Nova-building AGI unit announced the same day (count undisclosed). The pattern is now unmistakable: companies across the tech stack — infrastructure (Oracle), platform (monday.com), cloud (Amazon) — are simultaneously cutting staff to fund AI bets. The question is whether the AI products these reallocations fund will generate enough revenue to justify the human cost.
Trump Administration Commits $5 Billion to Genesis Mission for AI-Driven Science
The White House announced on July 22 more than $5 billion in federal commitments for the Genesis Mission, a whole-of-government AI-for-science initiative spanning 15+ agencies led by the Department of Energy. Secretary Chris Wright unveiled 278 project awards totaling $293 million (largest single project: $60 million for nuclear energy), drawing on 16 National Labs, 142 universities, and 157 companies. Arcee AI is among the industry partners building Genesis-Science-1, an American open-weight scientific model.
This is the federal government’s largest single commitment to AI research infrastructure. The $5 billion spans materials science, drug discovery, climate modeling, and nuclear energy — areas where AI has shown genuine acceleration potential but where private investment has lagged because the commercial returns are longer-horizon. The open-weight mandate for Genesis-Science-1 is particularly notable: it signals the administration believes frontier scientific AI should be publicly accessible, even as it restricts Chinese access to US commercial models. Fields medalist Terence Tao, incidentally, published a blog post the same day crediting ChatGPT with helping verify calculations in a major mathematics result — a small but telling data point about AI’s growing role in serious research.
That’s today’s wrap. The AMD-Anthropic deal is the story that will reverberate longest — it’s the first credible challenge to Nvidia’s monopoly on frontier AI training compute, and it comes with a $5 billion equity check attached. Alphabet’s cloud numbers prove AI infrastructure demand is still accelerating, and the White House accusation against Moonshot sets up the next chapter of the US-China AI rivalry. Follow NeuralPaws daily for the next AI news drop.