The Weekly AI Digest
Week of 4–13 May 2026
Top 5 AI Stories This Week
Musk Dissolves xAI, Hands 220,000 GPUs to Anthropic
On May 6, Elon Musk announced that xAI is being folded into SpaceX as a subsidiary called SpaceXAI. The same week, SpaceXAI signed a deal granting Anthropic exclusive access to the entire Colossus 1 supercomputer in Memphis: 300 megawatts and 220,000 NVIDIA GPUs. Anthropic immediately doubled Claude Code rate limits for Pro, Max, Team and Enterprise users and removed peak-hour restrictions. The deal is already online, unlike Anthropic's larger future commitments with AWS, Google and Microsoft.
Why it matters: Capacity constraints on Anthropic's flagship developer product have eased materially, which removes one of the longest-standing friction points for production Claude Code deployments. The deal also underscores how compute access has become the principal currency among frontier labs, with even commercial rivalries set aside to secure it.
Meta Commits Up to $100B to AMD for AI Chips
On May 5, Meta announced a 6-gigawatt agreement with AMD worth up to $100 billion, covering the MI450 GPU and next-generation EPYC CPUs. AMD issued Meta a performance-based warrant for up to 160 million shares (around 10% of the company) at $0.01 each, vesting against shipment milestones. First gigawatt of deployment begins H2 2026. Meta's projected 2026 AI capex is $135 billion.
Why it matters: AMD now has a credible second-source position to NVIDIA at hyperscaler scale, which has implications for chip pricing, supply availability and the inference cost curve over the next two cycles. The deal also confirms the pattern of hyperscalers taking equity stakes in their chip suppliers, reshaping the negotiating dynamic between AI compute buyers and sellers.
Anthropic Hits $30B ARR, Up 80x in 12 Months
Dario Amodei's May 11 investor letter put Anthropic's annualised revenue at $30 billion, up from $9 billion at the end of 2025 and roughly $400 million 12 months ago. Claude Code alone is now at $2.5 billion ARR and has more than doubled since January. Customers spending over $100K per year on Claude grew 7x year-over-year. Customers spending over $1M per year went from 12 two years ago to more than 500 today.
Why it matters: The $1M+ customer count is the more revealing number: enterprise Claude spend is now concentrated at scale, not distributed across a long tail. For commercial leaders, it signals that AI procurement at six and seven figures is becoming a normal range for organisations adopting frontier models seriously, not an outlier.
PwC to Deploy Claude Across Its Workforce
PwC announced on May 14 that it will deploy Claude across its global workforce. Amodei publicly named the scope as potentially "hundreds of thousands" of PwC employees. This is the largest single enterprise AI deployment announced by any frontier lab in 2026 and follows similar Anthropic moves with Lloyds Banking Group and Deloitte earlier in the year.
Why it matters: When a Big Four firm standardises on a frontier model at workforce scale, the implementation patterns flow through to its clients across audit, advisory and transformation engagements. PwC's choice of Claude becomes a procurement signal for every PwC client thinking about AI architecture in the next 12 months.
Coinbase Cuts 14% of Workforce Citing AI Restructuring
Coinbase announced a 14% workforce reduction (roughly 700 roles) on May 7, attributing the cuts directly to AI-driven productivity gains across engineering, support and operations. The pattern matches Meta, Microsoft, Snap and others reporting workforce contractions alongside record AI capex. Coinbase noted that AI now drafts the majority of incoming customer support responses and a growing share of engineering pull requests.
Why it matters: AI-attributed layoffs are now a recurring feature of quarterly results rather than an outlier story. The consistent picture across recent announcements is that AI productivity gains are being captured by employers in headcount reduction, with a structural impact on the entry-level and mid-career talent pipeline that will take years to surface.
Emerging Trends
Compute Is Consolidating Around Anthropic
The xAI-Anthropic GPU deal, combined with existing AWS, Google and Microsoft commitments, puts Anthropic in a compute position that no other frontier lab matches for immediately available capacity.
Anthropic's capacity constraints are easing faster than competitors'.
AMD Is Now a Credible NVIDIA Alternative
Meta's $100B AMD commitment and performance-linked warrant structure signals that the NVIDIA monopoly on frontier AI training is breaking. Enterprise procurement has a second option.
NVIDIA-only roadmaps are now a vendor concentration risk.
AI Layoffs Are Structural, Not Cyclical
Coinbase's 14% cut joins Meta, Microsoft and Snap in attributing workforce reduction directly to AI productivity. The pattern is consistent enough to treat as a baseline assumption.
Assume 10-15% structural reductions are achievable in knowledge work teams within 18 months.