Anthropic Surpasses OpenAI at $965 Billion Valuation Amid Revenue Surge
Anthropic’s $65 billion Series H round at a $965 billion post-money valuation has placed it ahead of OpenAI, whose most recent round valued the company at $852 billion. The financing, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia, nearly tripled Anthropic’s February valuation of $380 billion and incorporated $15 billion of previously committed capital, including $5 billion from Amazon.
Anthropic reported a $47 billion annualized revenue run rate, up from $30 billion earlier in the year and $10 billion in 2025, driven largely by adoption of its Claude Code coding assistant. The company also released Claude Opus 4.8 and previewed Claude Mythos, a model positioned for advanced cybersecurity workloads available only to select enterprise customers. These metrics illustrate how product-specific traction in developer tools can translate into rapid valuation gains even before broad profitability is achieved.
The valuation gap underscores a market that continues to reward execution on revenue-generating applications while remaining tolerant of heavy infrastructure spending. Anthropic’s trajectory now positions it to pursue an October listing potentially exceeding $1 trillion, intensifying pressure on OpenAI to demonstrate comparable commercial momentum before its own planned fourth-quarter debut.
OpenAI’s Resolution of the Erdős Unit-Distance Conjecture Demonstrates Scaled Reasoning
An internal OpenAI model produced a proof resolving the Erdős unit-distance conjecture, a discrete-geometry problem open since 1946. Fields Medalist Timothy Gowers described the result as “a milestone in AI mathematics,” while University of Toronto professor Daniel Litt called it “the first example of a result produced autonomously by an AI that I find exciting in itself.” The model synthesized techniques across multiple subfields rather than inventing new methods, after which human mathematicians refined and extended the argument.
This achievement follows a clear progression: three years ago, large language models struggled with basic arithmetic; last year they began dominating high-school competitions; this year they are contributing to long-standing research conjectures. The proof nevertheless relied on existing mathematical literature rather than novel conceptual breakthroughs, indicating that current systems excel at exhaustive exploration of known ideas but still require human insight to frame the most fertile questions.
The development points to a near-term division of labor in which AI systems supply breadth of recall and tolerance for low-success-rate search, while mathematicians retain advantage in depth and problem selection. Whether that division persists beyond the current decade remains uncertain given the pace of capability gains.
IPO Timing Uncertainties and Structural Losses Complicate OpenAI’s Path
Prediction markets currently assign a 30 percent probability that OpenAI’s IPO slips into 2027. The company filed its confidential S-1 on May 22 and targets a fourth-quarter listing at a valuation between $852 billion and $1 trillion. In contrast, SpaceX has already filed publicly with a planned June debut near $2 trillion, and Anthropic is preparing an October offering that could exceed $1 trillion.
OpenAI’s financial profile shows roughly $25 billion in annualized revenue alongside projected operating losses of $14 billion for 2026. Positive cash flow is not expected until around 2030. These figures stand in contrast to Anthropic’s $30 billion run rate and narrower path to profitability. Any delay in OpenAI’s listing would leave it last among the three largest private AI and space companies preparing to enter public markets, potentially affecting both employee retention and strategic optionality.
Election-Data Partnership Extends OpenAI’s Reach into Institutional Workflows
The Associated Press will supply OpenAI with certified U.S. election results for national, state, and local races through the 2028 general election. The arrangement gives ChatGPT users direct access to AP’s nonpartisan vote counts, which achieved 99.9 percent accuracy across nearly 7,000 races in 2024. OpenAI joins a roster of media, financial, and technology platforms that already rely on AP as the canonical source for election outcomes.
By embedding authoritative data inside widely used generative interfaces, the agreement reduces the surface area for misinformation while positioning OpenAI as infrastructure for civic information rather than merely a consumer chatbot. It also creates a template for similar integrations with other authoritative datasets, potentially expanding the model’s utility in regulated or high-stakes domains.
Ex-OpenAI Researcher’s Fund Signals Capital Reallocation Toward Compute Infrastructure
Situational Awareness, founded by former OpenAI researcher Leopold Aschenbrenner, disclosed a 5.6 percent stake in Nebius, a European cloud provider specializing in AI workloads. Nebius shares rose 7 percent on the news and stand 149 percent higher year-to-date. The company recently secured a $27 billion multi-year capacity agreement with Meta and a $2 billion investment from Nvidia, alongside a $2.6 billion fuel-cell deployment with Bloom Energy to address European power constraints.
Aschenbrenner’s fund has also taken positions in Oracle, Nvidia, ASML, and Micron, reflecting a thesis that physical bottlenecks—energy, chips, and data-center construction—will determine which organizations can sustain frontier model development. The allocation illustrates how capital is migrating from model-centric startups toward the specialized infrastructure layer required to train and serve next-generation systems.
Organizational and Marketing Realignments Reflect Intensifying Competitive Pressure
OpenAI has divided its top marketing responsibilities between two chief marketing officers, a structural change explicitly framed as preparation to counter rivals including Anthropic. The split occurs as both companies accelerate product releases and enterprise outreach, suggesting that scaling commercial operations now requires specialized leadership in distinct channels.
Taken together, the valuation reversal, mathematical proof, IPO timeline risks, election-data integration, and infrastructure investment patterns reveal an industry consolidating around a handful of well-capitalized players while simultaneously deepening dependence on specialized hardware and authoritative data sources. The next phase will likely be defined by which organizations most effectively combine frontier reasoning capabilities with reliable access to compute and trusted information pipelines.
