AI is congealing around a few choke points: Musk-linked compute, Nvidia/TSMC silicon, and frontier labs that now act like both utilities and VCs. At the same time those labs are soaking up capital, big tech is using AI as cover to cut tens of thousands of white-collar jobs and rebuild leaner org charts.
The money is in deciding how much of your future margin and data you’re willing to park inside that stack before model capabilities commoditize and the political blowback really lands.
Key Events
/Anthropic agreed to a compute deal with SpaceX worth up to $45B through 2029.
/Meta announced a $125B AI investment while simultaneously cutting thousands of jobs.
/OpenAI is offering $2M in API credits to every startup in the current Y Combinator batch in exchange for equity exposure.
/Intuit plans to lay off over 3,000 employees as part of its AI-focused restructuring.
/SpaceX’s IPO filing pegs its total addressable market at $28.5T, driven largely by AI opportunities.
Report
The live lever right now is AI infrastructure: who owns the pipes, who’s locked into them, and on what terms. Right behind it are AI‑driven org amputations and the frontier labs quietly turning into capital allocators and tollbooths.
musk–nvidia as de facto ai utilities
SpaceX is positioning itself as an AI compute utility, planning orbital data centers and tying AI growth directly to a $28.5T TAM in its IPO filing.
Anthropic’s cancellable multi‑year contract with SpaceX, valued up to $45B, makes Musk-linked infra a central dependency for one of the top labs.
SpaceX has accumulated over $37B in losses, so AI compute revenue from deals like Anthropic’s becomes a critical part of its profitability story.
In parallel, Nvidia has committed $90B to AI deals and expects $20B revenue from its Vera CPU, even as a China ban on the RTX 5090D V2 opens room for local competitors like Huawei and underscores geopolitical risk in the GPU supply chain.
Wafer shortages at TSMC and rising political and environmental pushback on data centers, including an EPA probe into a Meta facility over water contamination, cap how fast this infra can scale.
ai as a workforce shock, not just a tool
Meta cut around 8,000 roles, eliminated integrity and cybersecurity teams, and simultaneously announced a $125B AI investment, explicitly reframing headcount as fuel for automation.
Intuit is dropping more than 3,000 employees (about 17% of staff) while pushing hard into AI and online‑only services, with users already complaining about degraded support and product quality.
Oracle’s 30,000 layoffs and Amazon’s 16,000 cuts in 2023 were also attributed to AI, and Goldman Sachs estimates roughly 16,000 AI-related layoffs per month across the economy.
Salesforce says AI now handles 30–50% of its workload and has not hired a single software engineer since January 2025, tightening the definition of who counts as “core” talent.
On the other side, Gen Z is booing AI‑centric commencement speeches, explicitly linking AI to job loss and inequality, and CEOs are increasingly viewed as out of touch and profit‑obsessed when they hype AI while cutting staff.
labs turning into quasi‑VCs and tollbooths
OpenAI is prepping a confidential IPO with Goldman Sachs and Morgan Stanley while also offering $2M in API credits to every YC startup in the current batch in exchange for equity exposure.
Critics inside the developer community see this as a play to embed OpenAI deeply into startup architectures, creating high switching costs and homogenizing the LLM wrapper market.
Anthropic expects revenue to rise 130% in the June quarter. It projects revenue of $10.9B and later guides to an operating profit of $500M by Q2 2026, while also taking on a $1.5B piracy judgment that questions the durability of its balance sheet.
Salesforce plans to put roughly $300M into Anthropic tokens this year and has reorganized so aggressively around AI that it is being treated as a live case study in deep corporate dependence on a single lab.
At the same time, open-weight and regional models are catching up on many workloads, with Kimi 2.6 reportedly beating Gemini Flash while being ten times cheaper and Colorado’s SB051 explicitly carving out protections for open source, putting long‑term pressure on closed lab pricing power.
agentic and scientific ai are the new heavy workloads
A general-purpose OpenAI model autonomously solved the Erdős planar unit distance problem from 1946, marking the first time AI has cracked a prominent open problem in pure math.
Separate research claims AI can now perform in a week what would take a human researcher a year, and new benchmarks like Terminal‑Bench are being extended to cover full scientific workflows, shifting AI from code autocomplete to end‑to‑end R&D tooling.[Mathematics] Exa raised $250M to reorganize the web specifically for AI agents, signaling real capital flowing into infrastructure optimized for autonomous or semi‑autonomous systems.
Agentic systems are already in production in narrower domains: a med spa AI handles inbound calls and scheduling with humans only stepping in for complex cases, while Netflix uses multimodal AI to power richer video search.
At the same time, developers complain that AI‑generated code is often messy and hard to maintain, and generative agents are capable of autonomously writing exploits, reinforcing that these systems create new security and technical‑debt surfaces even as they unlock new workloads.
What This Means
Capital, power, and risk are concentrating around a small number of infra providers and labs at the exact moment AI is starting to bite into both infrastructure costs and white‑collar headcount. The main strategic decision is how much of your future margin and data you are willing to stake on those chokepoints versus building for portability in a world where core model capabilities are commoditizing but narratives, regulation, and public sentiment are not.
On Watch
/Nvidia’s RTX 5090D V2 ban in China creates an opening for domestic GPU vendors like Huawei and hints at a more fragmented global compute market.
/Colorado’s SB051 explicitly excluding open source from certain regulation could become a template for AI policy that structurally advantages open-weight models over closed APIs.
/The EPA’s investigation into a Meta data center over alleged water contamination may set new precedents for environmental constraints on AI-driven data center expansion.
Interesting
/Musk's xAI is under scrutiny for its safety culture as it approaches a $75 billion IPO, raising concerns about its operational practices.
/Polymarket has launched private company trading, allowing speculation on the future of Anthropic and OpenAI.
/Singapore's proactive AI investments and commitments from OpenAI position it as a leader in the global AI landscape.
/The competitive landscape for AI is intensifying, with Anthropic, OpenAI, and SpaceX all vying for limited investment capital.
/OpenAI's willingness to exchange $800M of compute for 2% equity in 400 startups reflects a bold investment strategy that could reshape the startup landscape.
We processed 10,000+ comments and posts to generate this report.
AI-generated content. Verify critical information independently.
/Anthropic agreed to a compute deal with SpaceX worth up to $45B through 2029.
/Meta announced a $125B AI investment while simultaneously cutting thousands of jobs.
/OpenAI is offering $2M in API credits to every startup in the current Y Combinator batch in exchange for equity exposure.
/Intuit plans to lay off over 3,000 employees as part of its AI-focused restructuring.
/SpaceX’s IPO filing pegs its total addressable market at $28.5T, driven largely by AI opportunities.
On Watch
/Nvidia’s RTX 5090D V2 ban in China creates an opening for domestic GPU vendors like Huawei and hints at a more fragmented global compute market.
/Colorado’s SB051 explicitly excluding open source from certain regulation could become a template for AI policy that structurally advantages open-weight models over closed APIs.
/The EPA’s investigation into a Meta data center over alleged water contamination may set new precedents for environmental constraints on AI-driven data center expansion.
Interesting
/Musk's xAI is under scrutiny for its safety culture as it approaches a $75 billion IPO, raising concerns about its operational practices.
/Polymarket has launched private company trading, allowing speculation on the future of Anthropic and OpenAI.
/Singapore's proactive AI investments and commitments from OpenAI position it as a leader in the global AI landscape.
/The competitive landscape for AI is intensifying, with Anthropic, OpenAI, and SpaceX all vying for limited investment capital.
/OpenAI's willingness to exchange $800M of compute for 2% equity in 400 startups reflects a bold investment strategy that could reshape the startup landscape.