Devin Hits $26B as AI Code Goes Enterprise-Grade — May 28, 2026

Note: Lighter post-holiday news cycle continuing. Three confirmed new stories in the strict 24-hour window (May 27–28). Anthropic $900B round remains ⚠️ Unconfirmed — no official announcement published. No stories duplicated from the May 25–27 briefing series.

⚡ Top Story

Devin Hits $26B: AI Coding Agent Goes From Demo to Defense Contractor

Cognition AI — makers of Devin, the autonomous AI software engineer — announced a $1 billion funding round on May 27 at a $26 billion post-money valuation. The round was co-led by Lux Capital, General Catalyst, and 8VC, with additional participation from Founders Fund, Ribbit Capital, Atreides, and Layer Global. Cognition's annualized revenue run-rate has reached $492 million, a 13-fold increase from $37 million in May 2025, driven by 50% month-over-month growth for six consecutive months.

The customer roster tells the story: Goldman Sachs, Mercedes-Benz, Citi, Dell, Santander, the US Army, and the US Navy are all running Devin in production. More striking: more than 90% of Cognition's own codebase is written by Devin.

Why it matters: Cognition's valuation trajectory — $4B (2024) → $10.2B (September 2025) → $26B (May 2026) — reflects a shift from AI-coding-as-demo to AI-coding-as-enterprise-infrastructure. The presence of US military customers isn't incidental; it signals that autonomous code agents have cleared security reviews and are operating in high-stakes production environments. If Cognition's 50% monthly growth trajectory continues even at half-speed, it will approach $1B ARR before year-end. The AI software engineer category has crossed from aspirational to billable.

Sources: Bloomberg — AI Coding Startup Cognition Raises $1 Billion at $26 Billion Value (May 27) · TechCrunch — Cognition raises $1B at $25B pre-money valuation (May 27) · The Decoder — Cognition doubles valuation to $26B in under nine months · The Next Web — 90% of code written by its AI

🔬 Research & Papers

No significant new AI research papers verified as published in the last 24 hours. ArXiv and Hugging Face Papers were inaccessible for real-time verification. Skip.

🏢 Industry & Startups

Sam Altman at CBA Conference, Sydney: "I'm Delighted to Be Wrong" on Jobs

Speaking at a Commonwealth Bank of Australia conference in Sydney on May 26, OpenAI CEO Sam Altman publicly walked back his earlier predictions about AI's impact on white-collar employment. Altman stated: "I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened." He added that he has personally experimented with using AI to respond to Slack messages and emails but returned to writing many himself, noting that "we really do care about our interactions with people."

Why it matters: Altman's admission is notable precisely because he was among the most prominent voices warning about near-term labor displacement. The correction revises the public record from one of the most influential figures in AI — and implicitly raises the question of why disruption is slower than expected: enterprise adoption friction, trust deficits, or limits in current models for ambiguous human-facing tasks. The "jobs apocalypse" narrative influenced regulatory conversations globally; this recalibration matters for policy as much as for workforce planning.

Sources: American Bazaar Online — Sam Altman says AI unlikely to lead to 'jobs apocalypse' (May 26) · IBTimes UK — 'I'm Delighted to Be Wrong' (May 26)

⚠️ Unconfirmed: Anthropic $900B Round May Have Formally Closed This Week

Anthropic's $30 billion funding round at a pre-money valuation above $900 billion — which Bloomberg (May 22) reported was expected to close "as soon as next week" — has not been confirmed by Anthropic as of publication. Multiple secondary sources use completed-tense language, but no official announcement has been verified. ⚠️ Watch: anthropic.com/news. Prior coverage: May 22, May 27 briefings.

🛠️ Tools & Releases

No new model releases, API updates, or open-source drops verified in the last 24 hours from frontier labs. Anthropic and OpenAI blogs returned access errors during research.

🌏 Global AI & Geopolitics

Qualcomm-ByteDance ASIC Deal: US Chip Commerce Finds a Path Through Export Controls

Qualcomm announced on May 26 a deal to supply millions of custom application-specific integrated circuits (ASICs) to ByteDance — owner of TikTok — for AI data center operations. The chips support ByteDance's AI agent software infrastructure and are engineered to comply with US export control limits by staying below the compute performance thresholds that restrict advanced GPU sales to China. ByteDance's AI infrastructure budget is approximately $29.4 billion (200 billion yuan). Qualcomm shares jumped 8% on the announcement.

The deal has two components: (1) bulk custom ASIC supply, and (2) manufacturing services where Qualcomm helps ByteDance bring its own proprietary chip design through production.

Why it matters: The deal establishes a template for how US semiconductor companies can maintain access to China's massive AI infrastructure market without triggering export control restrictions. Where Nvidia is effectively blocked from selling advanced GPUs to Chinese entities, Qualcomm is building custom ASICs engineered to the edge of regulatory thresholds. Whether this template holds under future export control tightening is the key risk — and the key question for every US chipmaker watching.

Sources: Bloomberg — Qualcomm Strikes AI Chip Deal with ByteDance for Data Centers (May 26) · WinBuzzer — Qualcomm's ByteDance Deal Expands Its AI Chip Push (May 27) · AI Weekly — Qualcomm Signs ByteDance ASIC Deal, Shares Jump 8%

⚡ Energy, Infrastructure & Chips

See 🌏 Global AI & Geopolitics above. The Qualcomm-ByteDance ASIC deal is the primary chip infrastructure story of the day.

🤖 AI Agents & Autonomy

Devin as Production-Grade Autonomous Agent — See ⚡ Top Story

Cognition's $26B raise and $492M ARR validate that fully autonomous coding agents are operating at enterprise scale in production environments, including defense procurement. More than 90% of Cognition's own codebase is now written by Devin. The US Army and US Navy are confirmed production users. This is no longer an agent demonstration — it is deployed agentic AI in high-stakes systems executing end-to-end software engineering workflows autonomously.

🔒 Safety, Alignment & Ethics

DataGrail 2026 Report: Most AI Vendors Don't Tell You What AI They're Using

DataGrail published its Privacy and AI Trends Report 2026 today (May 28), based on an analysis of 2,400 popular business software providers. Key finding: 63.6% of vendors that prominently advertise AI capabilities do not disclose a third-party AI subprocessor in their legal documentation — meaning enterprises purchasing AI-enabled software are likely exposing customer data to AI models they never reviewed, approved, or even know exist.

Additional findings:

32.8% of AI systems disclosing AI capabilities also disclose at least one other high-risk activity (sensitive data processing or automated decision-making)
Data deletion requests hit an all-time high for the fifth consecutive year — surging 567% since 2021, now representing 87% of all data subject requests
63% of audited websites still fail to comply with universal opt-out mechanisms (Global Privacy Control)

Why it matters: EU AI Act high-risk provisions take effect August 2, 2026 — 65 days away. Organizations using AI in healthcare, employment, credit, or law enforcement face mandatory transparency and documentation requirements. DataGrail's findings suggest most companies can't currently identify, let alone document, all the AI systems running in their vendor stack. This is a compliance gap measured in weeks, not years.

Source: VentureBeat — DataGrail report finds your vendor may be sending data to AI models you never approved (May 28)

📊 Numbers & Signals

$26B — Cognition post-money valuation (May 27, 2026) — from $10.2B in September 2025
$492M — Cognition annualized revenue run-rate — 13× growth from $37M in May 2025
50% — Cognition's MoM enterprise Devin usage growth, 6 consecutive months
90%+ — Share of Cognition's own codebase written by Devin
$1B — Cognition round size (co-led: Lux Capital, General Catalyst, 8VC)
8% — Qualcomm intraday stock gain on ByteDance ASIC deal
$29.4B — ByteDance AI infrastructure budget (200B yuan)
63.6% — AI vendors advertising AI that don't disclose AI subprocessors (DataGrail, May 28)
567% — Increase in data deletion requests since 2021 (DataGrail, 2026 report)
87% — Share of data subject requests that are now deletions (DataGrail)
65 days — Until EU AI Act high-risk AI provisions take effect (August 2, 2026)

🧠 Worth Thinking About

Three stories from the last 24 hours — Cognition's $26B raise, Sam Altman's Sydney mea culpa, and the Qualcomm-ByteDance deal — collectively describe the gap between what people predicted AI would do in 2026 and what it is actually doing. AI hasn't eliminated white-collar jobs at the rate Altman feared. But it has crossed $492M ARR selling to Goldman Sachs and the US Army, and is writing more than 90% of its own creator's codebase. Meanwhile, US chip policy designed to limit China's AI build-out is being navigated, legally and carefully, by a San Diego chipmaker supplying TikTok's data centers. The future arrived, but unevenly and laterally: not in the form of mass layoffs or a clean technological monopoly, but in the form of contracts, ASIC supply chains, and defense procurement. That's how AI is actually changing things right now — through billing cycles and procurement officers, not through headlines.

🏛️ Government & Regulation

No major new regulations, executive orders, or government AI initiatives announced in the last 24 hours.

Standing watch: EU AI Act high-risk provisions take effect August 2, 2026 (65 days). DataGrail findings above suggest most enterprises are not prepared.

🔭 Frontier Lab Dispatch

No confirmed posts from Anthropic, OpenAI, Google DeepMind, Meta AI, or leading Chinese labs published in the last 24 hours. Anthropic and OpenAI blogs returned access errors during research.

⚠️ Watch: If Anthropic formally confirms closure of its $30B+ round at $900B+ valuation, it will be published at anthropic.com/news and would supersede this briefing's top story.

🔗 Quick Links

Tier 3 — Tech & AI News Media

Tier 1 — Frontier Lab Watch

Anthropic Newsroom (watch for $900B round confirmation)