Europe and UK Rush to Build Sovereign AI

AI’s economics look unsustainable: hyperscalers and startups are burning cash while only hardware suppliers and data-center builders clearly profit. The author argues that AI won’t generate enough revenue to justify the massive capital expenditures hyperscalers have made — over $800 billion in the past three years and planned trillions more — citing Microsoft’s roughly $100 billion tied to its OpenAI partnership and an estimated $293.8 billion in capex since FY2023. Startups lose money, AI-specific revenues remain opaque, and claimed capacity (OpenAI’s 1.9GW) implies staggering build costs. The piece warns the AI bubble could pop unless business models and unit economics improve. Key players: Microsoft, OpenAI, NVIDIA.

Anthropic has agreed to buy the entire 300-megawatt compute capacity of xAI’s Colossus 1 data center in Tennessee for billions, after xAI reportedly shifted training to a newer Colossus 2 site. The deal highlights how compute is a strategic asset for frontier AI labs and raises questions about Grok’s actual compute needs and xAI’s capitalization strategy. Industry reaction sees the sale as unusual: leading AI firms typically hoard or vertically integrate compute rather than divest it, so selling a flagship facility suggests either excess capacity, financing pressure, or a different operational model from xAI and Elon Musk. The transfer could reshape capacity allocation, competition for GPUs, and bargaining power among top AI developers.

The article argues AI is currently uneconomic: hyperscalers and startups are burning vast sums on compute and data-center buildouts that won’t pay off. Citing testimony and analyst estimates, the author says Microsoft has spent roughly $100 billion tied to its OpenAI partnership and around $293.8 billion in capex since FY2023, with a large share supporting AI infrastructure. Hyperscalers reportedly invested over $800 billion in three years and plan another $1.7 trillion through 2027, implying they must generate trillions in AI revenue to break even. The piece warns that most AI startups lose money, AI revenue reporting is opaque, and the current investment wave risks an eventual market correction or “AI bubble” burst.

Two former OpenAI staffers and a coalition of AI safety nonprofits warned investors that xAI’s weak safety practices could pose “unpriced risks” for SpaceX ahead of its expected massive IPO. The critique, published as an investor letter by Guidelight AI Standards (cofounded by ex-OpenAI researchers Steven Adler and Page Hedley) and groups including Encode AI and The Midas Project, asks SpaceX to disclose whether xAI will keep developing frontier models and to publish public safety and governance plans. Signatories cite incidents involving xAI’s Grok chatbot — including violent and sexualized outputs and resulting regulatory scrutiny — and worry those liabilities could affect valuation, regulation, and litigation for SpaceX. The groups contrast xAI’s practices with peers like OpenAI, DeepMind and Anthropic, and urge greater transparency for investors.

Ramp’s dataset shows Anthropic now counts more business customers than OpenAI, a notable shift in enterprise AI adoption. The report compares customer sign-ups and usage across providers, highlighting Anthropic’s commercial traction—driven by its Claude models and enterprise-focused partnerships—while OpenAI remains dominant in broader market share and developer mindshare with GPT models. This matters because enterprise deals and integrations drive recurring revenue, influence product roadmaps, and shape which safety and privacy features businesses prioritize. The data suggests businesses may be favoring alternatives to OpenAI for contract terms, compliance, or model behavior. If sustained, the trend could accelerate competition among AI platforms, affect pricing and SLAs, and influence which models enterprises standardize on.

May 18, 2026 Economics LLM How China’s Shadow AI API Market Works China's shadow market offers access to Claude, Gemini, GPT, and other frontier models. Pay a local seller, get an API endpoint, connect… Read post

May 18, 2026 Economics LLM How China’s Shadow AI API Market Works China's shadow market offers access to Claude, Gemini, GPT, and other frontier models. Pay a local seller, get an API endpoint, connect… Read post

A rising U.S. backlash against AI is gaining attention after public comments from leading AI CEOs and waves of layoffs tied to automation narratives. Hacker News commenters point to executives’ rhetoric—especially from high-profile frontier labs—that generative models will replace jobs, fueling public anxiety beyond the tech sector. Complaints include overstated AI blame for pandemic-era overhiring, local impacts from data centers, rising costs for components like memory, and skepticism about job guarantees or UBI. The debate matters because corporate messaging, infrastructure impacts, and labor-market fears are shaping policy pressure, public trust, and the social license for AI development.

The Information analyzed 34 AI companies and found annual revenues approaching $80 billion, with combined year-over-year growth of 112% in six months and market share heavily concentrated: OpenAI and Anthropic together capture about 89% of reported revenue. Anthropic’s revenues have surged—reports suggest annualized revenue could reach $5 billion by June—while OpenAI has cited monthly revenue implying roughly $24 billion annualized. Anthropic records some income via cloud partners; OpenAI is contractually due to share roughly 20% of revenue with Microsoft through 2030. Several application-layer firms (Perplexity, ElevenLabs, Cognition, Cursor) have surpassed $50 million in annual sales but still pay substantial model-access fees to OpenAI and Anthropic, highlighting both rapid growth and dependency on the leading model providers.

OpenAI has reportedly announced that a country will become the world’s first to offer ChatGPT Plus free for all residents, according to the article’s Chinese headline. No further details are available in the provided material, including which country is involved, the start date, eligibility rules, funding arrangements, or whether access is delivered through government partnership, telecom bundles, or another program. If confirmed, a nationwide free ChatGPT Plus rollout would matter because it could accelerate consumer adoption of paid generative AI features, set a precedent for public-sector or national-level AI access programs, and influence competition among AI providers. More information is needed to verify the announcement and assess scope and implementation.

Anthropic has acquired API developer-infrastructure startup Stainless for a reported sum above €280 million, bringing Stainless’s SDK automation and API-to-SDK tooling in-house. Stainless, founded in 2022 by Alex Rattray, built a platform that auto-generates and updates SDKs across languages like Python, TypeScript, Kotlin, Go and Java, reducing integration and maintenance overhead for teams that depend on external APIs. Post-acquisition, Stainless will close its hosted services and stop providing official infrastructure updates to external customers, though existing generated codebases remain owned by those clients. The deal strategically gives Anthropic exclusive control over a connective layer used by AI labs, constraining competitors such as OpenAI and Google that previously relied on Stainless or similar services.

Anthropic has acquired the dev tools startup used by OpenAI, Google, and Cloudflare

OpenAI is hiring a Community Engagement Lead to reduce local opposition to its Stargate data center buildouts, signaling the company’s need to manage public perception as it expands compute infrastructure. The role—paying $129.6k–$236k plus equity—will liaise with communities near planned Stargate sites in Texas, Michigan, New Mexico, Wisconsin and Ohio, measure success by “reduced community resistance,” and integrate local concerns into planning. Stargate, a joint initiative with Oracle, SoftBank and MGX announced in 2025, aims to invest $500 billion in AI infrastructure across the U.S. The hire underscores tensions around water, energy, noise and temporary jobs tied to large-scale data centers, and reflects broader industry challenges securing local approvals for critical compute capacity.

A writer reports using Anthropic’s Claude to help navigate a consumer dispute and responds to reader pushback about relying on AI. They argue they did not blindly accept outputs but used the model as a tool—fact-checking, verifying sources, and applying critical judgment—while noting common objections: distrust of AI accuracy, privacy concerns, lack of transparency, and perceived skill erosion. The piece frames resistance as partly cultural and emotional (fear of change, status anxieties) and partly practical (overhyped claims, inconsistent model behavior, limited source traceability). It concludes that responsible AI use requires literacy, skepticism, and human oversight, and that broader adoption depends on trust, explainability, and better safeguards.

Anthropic’s Claude has overtaken OpenAI’s ChatGPT across several commercial and user metrics for the first time in years, including net new ARR, mobile downloads, business adoption, daily active users and annualized revenue. The shift reflects growing competition in the generative AI market as enterprises and consumers diversify beyond ChatGPT, favoring alternatives like Claude that emphasize safety, enterprise features, or different pricing and integrations. For OpenAI, slipping to second in these indicators could reshape partner strategies, product roadmaps and investor expectations. The change matters because it signals a maturing market where multiple models compete on performance, trust, platform support and go-to-market execution.

The Information : Analysis: 34 leading AI startups are generating ~$80B in annualized revenue, up 112% from six months ago, with Anthropic and OpenAI capturing 89% of the revenue — Anthropic and OpenAI are widening the revenue gap between themselves and the rest of the AI startup field.

Major AI providers have been intentionally subsidizing consumer and SMB subscriptions, selling powerful models at prices far below API economics to drive adoption. Companies like OpenAI, Anthropic, Google, Meta and newer entrants are absorbing heavy compute costs—examples include ChatGPT Plus and Claude Pro at $20/month while equivalent API usage would cost hundreds per user—creating a market distortion that lets enterprises build critical workflows on unsustainable pricing. The concern: providers will inevitably reprice or shift focus to enterprise contracts, exposing organizations to steep cost increases and operational risk. CTOs and finance leaders must model true token and compute usage now and renegotiate or architect for portability to avoid future bill shocks.