Move Over OpenAI: Why Anthropic’s $950B Valuation is Shaking Up the AI Arms Race

Anthropic's Cat Wu on the Future of AI: "Claude Will Anticipate Your Needs Before You Know What They Are":

Anthropic vs. OpenAI: The New Data Showing Who Really Rules the Enterprise Market:

In an exclusive interview at Code with Claude 2026, Anthropic's Head of Product for Claude Code and Cowork shares her vision for proactive AI, agent management, and why Anthropic never looks at its competitors.

The Moment: Anthropic Is Having Its Best Year Yet:

Few companies in the technology industry have had a year as consequential as Anthropic's 2026. As the AI arms race intensifies across every sector of the global economy, Anthropic is not just keeping pace — it is, by several measures, pulling ahead. The company is reportedly in talks to raise tens of billions of dollars in a new funding round that would place its valuation at approximately $950 billion, eclipsing OpenAI's March 2026 valuation of $854 billion.

Meanwhile, a recent market share report showed Anthropic has quadrupled its share of business customers since May 2025, outpacing OpenAI among enterprise users — the segment that increasingly defines who wins the AI platform wars.

At the centre of that enterprise momentum is Cat Wu, Anthropic's Head of Product for Claude Code and Cowork. Since joining Anthropic in August 2024, Wu has been a defining force in the company's product evolution — shepherding Claude's transformation from an informational chatbot into a full-featured coding agent and autonomous workflow platform. Frequently paired with Boris Cherny, the creator of Claude Code, Wu has been characterised within the industry as Anthropic's

"Batman and Robin" — a creative-technical partnership driving some of the most consequential product decisions in AI today. TechCrunch sat down with Wu at the second annual Code with Claude conference in San Francisco to discuss product philosophy, the future of human-agent collaboration, and what Anthropic is building toward next.

Product Philosophy: Why Anthropic Ignores Its Competitors:

In a competitive landscape defined by rapid feature releases, copycat launches, and constant benchmarking against rivals, Cat Wu's product philosophy stands out for its deliberate rejection of that entire framework. When asked how much of Anthropic's product strategy is shaped by watching competitors, Wu's answer was unambiguous.

"If you think about competitors, you end up being perpetually two weeks, or a month behind how fast you can execute. It's normally not the best way to stay at the frontier." For Wu and her team, the guiding principle is what she calls "staying on the exponential." The premise is straightforward but demanding: AI capability is improving on a consistent upward curve, and the job of Anthropic's product team is to ensure that Claude tracks that curve as closely as possible.

Following a competitor's roadmap, by definition, means trailing the frontier rather than defining it. This philosophy explains why Anthropic has maintained a release cadence that few companies in any industry can match — at least six new models in 2025 alone, with nearly as many already released in the first half of 2026.

The pace raises an obvious question: can it continue? Wu's answer is both honest and revealing. "Our hope is that it continues," she said with a laugh — acknowledging the ambition while signalling that the rate of model improvement genuinely justifies the release frequency. The models are getting better. The deployments, however, are getting more nuanced — a point illustrated by Anthropic's handling of its most sensitive AI release to date.

Glasswing and Mythos: When AI Becomes Too Powerful to Release Publicly:

Not every breakthrough gets a public launch — and Anthropic's handling of its Glasswing initiative is perhaps the clearest signal yet of how the company thinks about AI risk at the frontier. Launched in April 2026, Glasswing is a controlled-access programme that gave a small consortium of partner organisations — including Amazon, Apple, CrowdStrike, and Microsoft — exclusive early access to Mythos, Anthropic's new cybersecurity AI model.

Mythos is designed to scan codebases for software vulnerabilities — a capability with enormous positive applications for enterprise security teams. But Anthropic made the unusual decision not to give it a general public release. The stated reason: the model is too powerful and too easily weaponised.

A tool that can find security vulnerabilities at scale can, in the wrong hands, be used to exploit them at scale. Rather than withhold the capability entirely, Anthropic opted for a curated deployment — placing Mythos in the hands of organisations with the security infrastructure and accountability to use it responsibly.

For Wu, Glasswing represents the kind of deployment nuance that becomes increasingly necessary as AI models grow more capable. " As much as possible, we want this intelligence to benefit as many people as possible," she said, "and it has to be handled in a very safe way." The Glasswing model — selective, accountable, partner-gated access for high-risk AI capabilities — may well become the template for how the most powerful future models are deployed.

Managing AI Agents: The New Leadership Skill Nobody Taught You:

The most provocative part of Cat Wu's worldview is also, arguably, the most consequential for the future of work. Wu has said publicly that she believes the future of professional work is humans managing fleets of AI agents — a vision that raises immediate questions about what happens when those agents become more capable than the humans overseeing them.

"I think it is extremely hard to manage agents if you can't do the job yourself. The managers still need to be experts in their domain."

Wu's position on this is nuanced and worth sitting with. She does not believe that increasing AI capability diminishes the need for human expertise. Quite the opposite: she argues that managing AI agents effectively requires deeper domain knowledge, not less. "You have to understand, why did the agent make this mistake? Did it misinterpret my instruction? Was my request under-specified? You have to have the ability to debug it." This is a fundamentally different skill set from traditional task execution — closer to the work of a senior manager or technical lead than a junior contributor.

The parallel to people management is apt and instructive. Good managers do not simply delegate and receive results — they understand the work well enough to identify where and why things go wrong. In the age of AI agents, that same principle applies. The professionals who will thrive in an agentic workplace are not those who know how to use AI tools, but those who know their domain deeply enough to audit, correct, and direct AI outputs with genuine authority.

Jobs, Teams, and the Question Everyone Is Asking:

The conversation inevitably turns to the question that sits beneath every discussion of agentic AI: what happens to the humans? If AI agents can handle the work of an intern, a junior analyst, or an entry-level engineer, does that mean organisations simply stop hiring those roles?

Wu's answer is optimistic — and notably personal. "For me, it's responding to emails. I think everyone has this part of their life." Her vision is not of mass replacement, but of mass liberation from the tedious, repetitive, low-satisfaction tasks that consume a disproportionate share of most professionals' working days. The ambition, as she frames it, is that AI handles the work people dread — so that people can spend more time on the work that genuinely energises and fulfils them.

Whether that vision plays out at the level of the individual or the organisation is a question Wu does not fully resolve — and to her credit, she does not pretend to. The honest reality is that agentic AI will compress team sizes in some functions, even as it expands output per person. The net effect on employment, economic opportunity, and workplace culture is one of the defining questions of the next decade — and one that product leaders at companies like Anthropic will play a central role in shaping, whether they seek that role or not.

The Next Big Thing: Proactive AI That Acts Before You Ask:

So what is Anthropic building toward in the next six months? Wu's answer points to a shift that may be the most significant change in how people experience AI since the launch of ChatGPT: proactivity.

To understand the significance of that word, it helps to trace the arc Wu describes. The first era of modern AI was synchronous — you asked a question, the AI answered. The current era is routine-based automation — AI handles defined, repeating tasks like responding to customer support tickets on a schedule. The next era, as Wu sees it, is something qualitatively different: Claude understanding what you work on and proactively setting up automations for you — without being asked.

"The next step is that Claude understands what you work on, and just sets up some of these automations for you." — Cat Wu, Head of Product, Claude Code & Cowork This is the vision of AI as a genuinely intelligent colleague, not a tool.

A tool waits to be used. An intelligent colleague observes, anticipates, and acts — bringing you the information you need before you realise you need it, drafting the document before you open a new page, flagging the risk before it becomes a problem. Proactive AI represents the moment Claude stops being reactive and starts being genuinely assistive in the fullest sense of that word.

For enterprise users, the implications are substantial. Proactive AI agents that understand an organisation's workflows, data, and priorities could fundamentally change the economics of knowledge work — reducing the overhead of coordination, communication, and status management that currently consumes an outsized share of professional time. If Anthropic can deliver on this vision in the next six months, it will represent a genuine step-change in what AI can do for businesses.

Key Takeaways: Cat Wu's Vision for Claude and Anthropic:

• Anthropic ignores competitors by design. Its product strategy is built around tracking the AI capability curve, not reacting to rival feature releases.

• Anthropic is approaching a $950B valuation and has quadrupled business customer market share since May 2025, outpacing OpenAI in the enterprise.

• Glasswing and Mythos signal a new deployment model for high-risk AI: curated, partner-gated access rather than general public release.

• Managing AI agents requires deep domain expertise. The professionals who thrive will be those who can audit, correct, and direct AI outputs — not just use AI tools.

• Proactivity is the next frontier: Claude will move from answering questions to anticipating needs — setting up automations before you ask.

• Wu's optimistic take on jobs: AI handles the tedious, repetitive work so professionals can focus on high-value, high-satisfaction tasks.

The Bottom Line: A Product Leader Who Thinks in Decades:

What makes Cat Wu's perspective valuable is not just her proximity to Anthropic's product decisions — it is the clarity and consistency of her underlying framework. Stay on the exponential. Build for depth, not breadth.

Manage AI agents the way you manage people: with expertise, accountability, and the willingness to debug when things go wrong. And above all, design for a future where AI does not replace human ambition — it amplifies it.

That vision of proactive, anticipatory AI is closer than it might seem. And if Anthropic delivers on it, the experience of working with Claude six months from now may feel less like using software and more like having a genuinely intelligent, context-aware partner —

one that understands your work, your priorities, and your needs well enough to act on them before you have to ask.

Published May 14, 2026 | AI Leadership | Product Strategy | Claude AI | 10 min read