Anthropic Overtakes OpenAI, Anduril's $61B Defense Play, and Python's Great GC Reversal

The tech landscape this week delivered a sequence of developments that, taken together, paint a picture of an industry simultaneously racing forward and doubling back to fix its foundations. Anthropic has quietly overtaken OpenAI in enterprise adoption, a $5 billion defense tech round rewrote the fundraising record books, a geothermal startup rode the AI data center wave to a blockbuster IPO, Python's core team made a rare and humbling reversal, and Microsoft found itself in the crosshairs of a prolific zero-day leaker. Here's a deep dive into the stories that matter.

Anthropic Overtakes OpenAI in Enterprise Adoption

For the first time since the generative AI boom began, Anthropic has more verified business customers than OpenAI. According to the latest AI Index from fintech firm Ramp, which surveys expense data from over 50,000 companies, 34.4% of businesses are now paying for Anthropic services compared to 32.3% for OpenAI. The shift is especially pronounced in high-adoption sectors like finance, technology, and professional services.

The numbers tell a dramatic story of momentum. In May 2025, only 9% of businesses were paying for Anthropic products. Over the following 12 months, that figure rocketed by 26 percentage points while OpenAI's share actually declined by 1%. Ramp economist Ara Kharazian noted that Anthropic had already been leading among the highest-adoption groups, and the trend has only accelerated.

This isn't just a vanity metric. Enterprise AI contracts are the single most valuable revenue stream in the industry, and Anthropic's surge suggests that the market is diversifying beyond a single dominant provider. Companies are increasingly running multi-model strategies, deploying Claude for sensitive workflows while maintaining GPT for others. For a sector that many feared was becoming a monopoly, these numbers are encouraging. The enterprise AI race, it turns out, is far from over.

Anduril Raises $5B at $61B Valuation as Defense Tech Scales

Defense technology company Anduril has closed a $5 billion Series H funding round at a $61 billion valuation, led by returning investors Thrive Capital and Andreessen Horowitz. The round more than doubles Anduril's valuation from $30.5 billion, where it stood just under a year ago after raising $2.5 billion led by Founders Fund. CEO Brian Schimpf announced the raise alongside news that the nine-year-old company doubled its revenue in 2025 to $2.2 billion.

The capital injection comes at a moment when defense tech is transitioning from speculative venture bet to established industry category. Anduril has secured major contracts including a role in developing a space-based "golden dome" missile defense shield for the United States and a contract with the Netherlands' Ministry of Defence. The company's trajectory mirrors a broader pattern: governments worldwide are increasingly willing to partner with startups for critical defense capabilities, a departure from the traditional reliance on legacy contractors.

What makes this round particularly significant is the sheer scale. A $5 billion single round places Anduril among the largest venture capital raises in history, and the $61 billion valuation positions it as one of the most valuable private companies globally. For the defense tech sector, this is a watershed moment that signals institutional investor confidence is no longer tentative.

Fervo Energy IPO Rides AI Power Demand to $10B Valuation

Geothermal startup Fervo Energy made a triumphant public debut on the Nasdaq, with its stock popping 33% on its first trading day under the ticker FRVO. The company raised $1.89 billion in an upsized IPO, pricing shares at $27 after lifting the price range twice to meet investor demand. The strong debut pushed Fervo's market valuation past $10 billion.

The driving force behind Fervo's IPO success is the same force reshaping the entire energy sector: artificial intelligence. Data centers powering AI workloads require enormous, reliable baseload electricity, and geothermal energy offers exactly that. Google has already committed to purchasing 115 megawatts from Fervo's Corsac Station project in Nevada. As tech giants scramble to secure clean, around-the-clock power for their AI infrastructure, geothermal has emerged as one of the most compelling solutions.

Fervo's IPO is more than a single company's success story — it represents the growing financial market recognition that the AI boom is fundamentally an energy story. As we've explored previously, the gap between AI compute ambitions and available power supply is one of the industry's most critical bottlenecks. Fervo's public market reception suggests investors are willing to bet big on closing that gap.

Python Reverts Its Incremental GC in a Rare Humbling Moment

In what has become a significant event in the Python ecosystem, core developers have decided to revert the incremental garbage collector introduced in Python 3.14 after receiving reports of substantial memory pressure in production environments. The revert will ship in an expedited patch release (Python 3.14.5) and will also apply to the still-in-alpha Python 3.15, returning the garbage collector to the proven generational implementation from Python 3.13.

The decision carries important lessons for the broader developer community. The incremental GC was rolled out without a formal PEP (Python Enhancement Proposal) process, a departure from Python's typically rigorous standards for changes to core subsystems. Core developer Hugo van Kemenade acknowledged this in the discussion, noting that if the team wants to reintroduce incremental garbage collection for Python 3.16, it must go through the standard PEP process and receive more thorough evaluation.

This is a textbook example of why "move fast and break things" doesn't translate well to foundational infrastructure. Python powers everything from web frameworks to machine learning pipelines, and a regression in memory management can cascade across millions of deployments. The Python team's willingness to openly acknowledge the mistake and revert quickly, rather than doubling down, is a model for responsible open source governance. It also highlights a growing tension in language ecosystems: the pressure to innovate versus the need to protect the stability that made the platform dominant in the first place.

Two New Microsoft Zero-Days Released After Patch Tuesday

A security researcher operating under the aliases Nightmare-Eclipse and Chaotic Eclipse has publicly disclosed two new unpatched Microsoft zero-day vulnerabilities, releasing them just after the company's monthly Patch Tuesday update. Dubbed YellowKey and GreenPlasma, the vulnerabilities represent serious threats to Windows systems.

YellowKey is a BitLocker bypass that allows an attacker with physical access to gain an unrestricted shell via a USB drive, using the correct key sequence without needing a BitLocker PIN. This is particularly alarming for enterprise environments that rely on BitLocker as a last line of defense for stolen or misplaced laptops. GreenPlasma, meanwhile, is a privilege escalation flaw that grants SYSTEM-level access to attackers — and currently has no available mitigation.

These are the fourth and fifth zero-days disclosed by this researcher this year alone, following the earlier release of RedSun and UnDefend, which targeted Windows Defender. Three of the five remain unpatched. The pattern is deeply concerning: a single individual is systematically exposing flaws in Microsoft's security architecture, and the company's patch cycle is unable to keep pace. As we've seen with similar disclosures in the open source world, when researchers start bypassing coordinated disclosure, it usually signals deeper frustration with the vendor's response times and processes. For enterprise security teams, this means the gap between known vulnerabilities and available patches is wider than it should be.

The Bigger Picture

What connects these five stories is a theme of recalibration. Anthropic's enterprise surge shows the AI market recalibrating away from a single-provider mindset. Anduril's massive raise reflects defense spending recalibrating toward startups. Fervo's IPO shows energy markets recalibrating to accommodate AI's voracious power demands. Python's GC revert is the language itself recalibrating to protect its stability foundation. And the Microsoft zero-days expose the ongoing need for the security industry to recalibrate its response to increasingly prolific vulnerability disclosure.

In a week that lacked a single defining headline, the through-line was clear: across AI, defense, energy, developer tools, and cybersecurity, the tech industry is in a period of structural realignment. The organizations that thrive will be those that recognize these shifts early and adapt — not just the ones moving fastest, but the ones building on the most solid ground.