News
Anthropic Hits $965B Valuation, Tops OpenAI as Mythos Looms
<p>Anthropic is now the most valuable AI startup on the planet. The San Francisco lab said Thursday it raised <strong>$65 billion</strong> in a new round that values the company at <strong>$965 billion</strong>, pushing it past OpenAI and to the edge of a trillion-dollar line no private AI company has crossed. The same day, it shipped Claude Opus 4.8, a model it says is less willing to deceive users or help with misuse than the version before it.</p>
<p>The celebration came wrapped in an awkward second headline. In the very announcement that crowned Anthropic the safety-first leader of the industry, the company said it will bring its Mythos-class models, built to hunt and chain software vulnerabilities, to all customers &#8220;in the coming weeks.&#8221;</p>
<h2>The $65 Billion Round, Broken Down</h2>
<p>The Series H was led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital, with Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ and XN co-leading. It is the kind of investor roster that shows up only when a private listing is close.</p>
<p>Of the headline figure, roughly $15 billion was previously committed capital from hyperscalers, the cloud giants that rent out the computing power these models run on. That tranche includes <strong>$5 billion from Amazon</strong>, which means close to $50 billion was freshly written in this round alone. In February, the company&#8217;s Series G valued it at $380 billion. The price has roughly two and a half times in about three months.</p>
<p>The new number does more than set a record. It rearranges the pecking order at the top of the industry.</p>
<table>
<thead>
<tr>
<th>Company</th>
<th>Latest valuation</th>
<th>Date set</th>
<th>Round size</th>
</tr>
</thead>
<tbody>
<tr>
<td>Anthropic</td>
<td>$965 billion</td>
<td>May 2026</td>
<td>$65 billion (Series H)</td>
</tr>
<tr>
<td>OpenAI</td>
<td>$852 billion</td>
<td>March 2026</td>
<td>$122 billion</td>
</tr>
</tbody>
</table>
<p>OpenAI raised far more cash in March, $122 billion against Anthropic&#8217;s $65 billion, yet sits at a lower headline valuation. You can read more in <a href="https://openai.com/index/accelerating-the-next-phase-ai/" target="_blank" rel="noopener">OpenAI&#8217;s own account of its $122 billion raise</a>. The gap says investors are now pricing Anthropic&#8217;s growth curve, not just its cash pile.</p>
<figure class="wp-block-image aligncenter featured-image" style="margin:1.5em auto;text-align:center;"><img class="aligncenter" src="https://budgyapp.com/wp-content/uploads/2026/05/anthropic-965-billion-valuation-tops-openai-ahead-of-mythos-ai-model-release.webp" alt="Anthropic $965 billion valuation tops OpenAI ahead of Mythos AI model release." style="width:100%;max-width:800px;height:auto;border-radius:8px;display:block;margin:0 auto;" /><figcaption style="text-align:center;font-size:0.85em;color:#888;margin-top:0.5em;">Anthropic $965 billion valuation tops OpenAI ahead of Mythos AI model release.</figcaption></figure>
<h2>Why Investors Paid Up: A $47 Billion Run-Rate</h2>
<p>The growth is the argument. Anthropic said its annualized run-rate revenue, the figure you get by taking current monthly sales and projecting them across a full year, crossed $47 billion by May. That is up from about $10 billion at the close of 2025, a jump of roughly 4.7 times in five months.</p>
<ul>
<li><strong>$47 billion</strong> annualized run-rate revenue, reported in May 2026</li>
<li><strong>$10 billion</strong> run-rate at the end of 2025</li>
<li>Roughly <strong>4.7x</strong> growth in about five months</li>
<li>Run-rate had touched only about $14 billion at the February Series G</li>
</ul>
<p>For scale, OpenAI has said it generates around $2 billion a month, which annualizes near $24 billion. On that measure Anthropic is now running ahead of the company that defined the consumer chatbot era. Most of that lead is not a consumer story. It is enterprise software spend, led by Claude Code, the agentic coding tool businesses have adopted at speed, alongside a surge in paying users. You can track the model family on <a href="https://www.anthropic.com/claude" target="_blank" rel="noopener">Anthropic&#8217;s Claude product page</a>.</p>
<h2>Mythos Arrives With a Cyberweapon&#8217;s Resume</h2>
<p>Here is where the safety badge gets heavy. The Mythos-class system, whose existence Fortune first reported in March after a data leak, is built around offensive cyber skill. In Anthropic&#8217;s own red-team testing, it found and exploited unpatched flaws across major operating systems and web browsers when a user pointed it at a target.</p>
<p>The capabilities go further than spotting bugs. According to Anthropic&#8217;s published preview, the model can discover unknown vulnerabilities, write working exploits, and chain them together into an attack sequence, the sort of work that normally takes a skilled human team.</p>
<ul>
<li>Finds zero-day flaws, weaknesses no one has patched yet, in widely used software</li>
<li>Builds complete, working exploits, even when the operator has no security training</li>
<li>Chains separate vulnerabilities into a single sophisticated attack</li>
<li>Reconstructs plausible source code to target closed software it cannot see</li>
</ul>
<p>The danger is not abstract. Anthropic has warned the tool could let attackers compromise the networks behind critical infrastructure, from banking systems to electric grids. That risk is already moving through official channels: the company has agreed to brief regulators on banking-software flaws its preview model surfaced, a story covered in detail in <a href="https://budgyapp.com/anthropic-mythos-fsb-cyber-briefing/" target="_blank" rel="noopener">Anthropic&#8217;s planned Financial Stability Board cyber briefing</a>. The technical specifics sit in <a href="https://red.anthropic.com/2026/mythos-preview/" target="_blank" rel="noopener">Anthropic&#8217;s Claude Mythos Preview disclosure</a>.</p>
<p>For now, access is gated. Through a limited program, only a select set of partners can use the model, and chiefly to find and patch holes in their own systems before anyone else does.</p>
<h2>The Safety Pitch Meets Its Hardest Test</h2>
<p>Anthropic has built its brand on restraint, and Opus 4.8 is the company leaning into that story, pitched as harder to trick into deception or misuse. So the plan to widen Mythos access lands as a genuine tension rather than a talking point.</p>
<p>The company says it will not release the model broadly until it has built safeguards that make malicious use meaningfully harder. That promise is the whole ballgame. A safety-first lab releasing the most capable offensive hacking tool yet built only works if the guardrails hold under real-world pressure, against users who do not share Anthropic&#8217;s intentions.</p>
<p>There is a counterweight worth stating plainly. Several cybersecurity researchers have told reporters that capable existing models, including ones from Anthropic and OpenAI, can already surface many of the same flaws, and that defenders gain from finding them first. The fear and the upside live in the same feature. Whether the &#8220;coming weeks&#8221; timeline survives contact with that debate is the live question.</p>
<h2>The OpenAI Scoreboard Just Flipped</h2>
<p>For two years the valuation race had a clear leader, and it was not Anthropic. OpenAI set the pace, raised the biggest rounds, and owned the consumer mindshare through ChatGPT. The March round at $852 billion looked like another lap won.</p>
<p>Two months later the order is reversed on both valuation and reported run-rate. Anthropic&#8217;s enterprise-heavy model, less flashy than a 900-million-user chatbot, has turned into the faster revenue engine, a contrast to the reasoning-model arms race that produced systems like <a href="https://budgyapp.com/openai-o1-model-thinks-reasons-breakthrough/" target="_blank" rel="noopener">OpenAI&#8217;s step-by-step o1 reasoning models</a>. Full announcements sit on <a href="https://www.anthropic.com/news" target="_blank" rel="noopener">Anthropic&#8217;s company news page</a>.</p>
<h2>The IPO That Everyone Sees Coming</h2>
<p>All of this is happening as Anthropic moves toward one of the most anticipated public listings of the year. The company is reportedly preparing for an initial public offering (IPO, the first sale of shares to public investors), with some accounts pointing to an October window, even as it keeps raising at record private prices.</p>
<p>The Series H reads like a last private round before that step. A roster of crossover funds, a near-trillion-dollar mark, and a revenue line growing faster than almost any company its size has ever managed are exactly the ingredients a banker wants on the cover of a prospectus.</p>
<p>The risk is that the two halves of Thursday&#8217;s news pull against each other once public-market scrutiny arrives. A $965 billion price tag invites questions about how durable a 4.7-times revenue jump really is, and a model that can break into critical systems invites questions no other software IPO has had to answer.</p>
<p>If the safeguards hold and the run-rate keeps climbing, Anthropic prints the listing on its own terms. If Mythos reaches the wider world before the guardrails are ready, the most valuable AI startup becomes the most scrutinized one instead.</p>