Anthropic hits $45B ARR, chasing a $900B valuation — making it the most valuable AI company on earth
Anthropic is reviewing investor offers for a new funding round of up to $50 billion at a pre-money valuation of roughly $900 billion, per the Financial Times. That would surpass OpenAI’s March valuation of $852 billion. The company’s annualized revenue is now approaching $45 billion — a fivefold increase from $9 billion at end of 2024.
Why this matters: This is the fastest revenue ramp in software history: from $87M ARR in January 2024 to $45B today. In 28 months. The headline story is Claude, but the real engine is Claude Code — a product that didn’t exist 18 months ago and now generates over $1B ARR on its own. An IPO as early as late 2026 is on the table.
Meta to lay off 8,000 starting May 20 — the AI restructuring playbook in action
Meta will begin laying off approximately 8,000 employees on May 20, representing 10% of its 78,865-person global workforce, with more waves expected through H2 2026. Despite posting $200B revenue and $60B profit in 2025, Meta is reallocating payroll toward a $115-135B AI infrastructure bill. Employees are being reorganized into “AI pods” under Chief AI Officer Alexandr Wang’s new Superintelligence Labs division.
Why this matters: This isn’t a financial rescue — it’s a deliberate restructuring. Zuckerberg’s bet is that small AI-augmented teams can deliver what entire departments once produced. Since late 2022, Meta has cut roughly 25,000 workers. The new job families at Meta? AI Builders, AI Pod Leads, and AI Org Heads.
NZ angle: For Kiwis in Meta’s亚太 operations or NZ-based remote contractors, the impact hasn’t been specified at the country level yet. But the wider signal — that Big Tech sees AI as a headcount replacement strategy — should be worrying for NZ’s growing tech workforce.
Google ends whiteboard coding interviews — allows AI assistants
Google is piloting a new interview format allowing software engineering candidates to use AI assistants during a “code comprehension” round, initially for junior to mid-level US roles. The company says it reflects “how our teams are operating in the AI era.”
Why this matters: This ends a foundational assumption of elite tech hiring — that unassisted, whiteboard-style coding ability is the right signal for engineering talent. Two camps have formed: those who see it as intellectually honest (you use AI on the job), and those worried it collapses the ability to distinguish strong engineers from strong prompt writers. Google is giving the rest of the industry permission to follow — expect AI-assisted interview rounds to be standard within 18 months.
OpenAI launches Codex Chrome extension — AI that acts on websites
OpenAI released a Chrome extension for Codex that performs tasks directly on websites including LinkedIn and Salesforce, part of a “super app” push combining ChatGPT, Codex, browsing, and agentic capabilities into a single cross-platform system.
Why this matters: This is the inflection from “AI that helps you code” to “AI that does things on the web for you.” The browser is where most work happens — and OpenAI just moved in. For enterprise IT and security teams, this is a new category of risk: AI extensions that can take actions on authenticated websites.
EU reaches provisional deal to simplify AI Act — delays high-risk rules
The Council of the EU and European Parliament agreed on May 7 to targeted amendments to the AI Act as part of the Digital Omnibus package. High-risk AI rules will be delayed until 2027/28, and the “Omnibus VII” package aims to streamline overlapping digital regulations. The deal clarifies where AI rules overlap with machinery regulations.
Why this matters: The EU is walking a tightrope — simplifying rules to keep industry competitive while still maintaining a regulatory framework. The delay on high-risk rules gives companies breathing room but also signals that full enforcement is still years away. This is a pragmatic shift from the original maximalist approach.
Google “Remy” and Meta “Hatch” — the personal AI agent race heats up
Google is testing “Remy,” a 24/7 personal AI agent inside Gemini that connects to Google services. Meta is developing “Hatch” along with an agentic shopping tool for Instagram, trained in sandboxed environments simulating real websites. Both are direct responses to Anthropic’s Claude Code/Claude Cowork and OpenAI’s agent push.
Why this matters: Google shut down its browser agent Project Mariner on May 4 to focus on Remy. Meta tried to acquire Chinese agent startup Manus for $2B but China’s regulators blocked it. The browser agent wave is fading — the market is shifting toward integrated personal agents that live inside email, calendars, office tools, and shopping platforms.
Anthropic and OpenAI move into services — buying consulting firms
Joint ventures tied to both companies are in talks to acquire services firms that help businesses deploy AI. OpenAI’s venture is in advanced stages on three deals. Anthropic separately announced a new enterprise AI services company backed by Blackstone, Hellman & Friedman, and Goldman Sachs.
Why this matters: Model providers want to be in the driver’s seat for enterprise AI implementation, not just supply the technology. For CIOs, this means lower deployment risk but deeper vendor lock-in. AI model providers are repositioning as “one-stop shops” for enterprise transformation, competing directly with Accenture, Deloitte, and the legacy systems integrators.
AI data centres on track to consume 1,000 TWh by 2026 — grid crisis looming
Global data centre electricity consumption is projected to exceed 1,000 TWh by 2026 as AI workloads scale, according to the IEA and Bloom Energy’s 2026 Power Report. That’s roughly equivalent to Japan’s entire national electricity use. Some EU member states have imposed moratoriums on new data centre construction in energy-constrained regions.
Why this matters: The grid can’t keep up. Interconnection timelines are widening, and AI’s power demand is creating systemic risk. This is the under-appreciated bottleneck on AI’s growth — not compute, not models, but power.
Australian residents fight AI data centre boom — and NZ should watch closely
Residents in West Footscray, Melbourne are pushing back against NextDC’s M3 “hyperscale AI factory” — a 10-hectare complex drawing 225MW of power with 100 diesel generators. Similar fights are happening in Sydney’s Lane Cove and across three states. Residents argue planning approvals didn’t account for the scale of AI data centres.
Why this matters for NZ: As NZ explores its own data centre expansion (Microsoft, AWS, and local players), the Australian backlash is a preview. NZ’s Resource Management Act and local council processes will face similar tensions between AI infrastructure needs and community impact. The AI Blueprint for Aotearoa’s new “Sustainable AI” workstream is already convening conversations from April to June 2026.
Singapore’s PM Wong says NZ and Singapore can lead AI deployment together
Speaking at the Singapore-NZ Leadership Forum, PM Lawrence Wong said both countries are unlikely to build the largest foundation models but can succeed by finding a niche in AI deployment. This follows a Parliamentary committee report warning that large international businesses “are not yet getting a return from AI” — but supporting NZ firms in pursuing it anyway.
Why this matters: This is a realistic framing. NZ doesn’t need to build the next GPT. But specialised AI deployment — agritech, climate, creative industries — is where small, agile economies can compete. The challenge is whether NZ has the talent pipeline and regulatory clarity to actually execute.
🔍 THE BOTTOM LINE
This afternoon’s news stack says something clear: the AI industry is consolidating fast. Anthropic is worth nearly a trillion dollars in under three years. Meta is cutting 8,000 people to fund $135B in AI infrastructure. Google is rewriting its hiring playbook. The EU is softening its rules. And the grid that powers all of this is starting to creak. The phase of “move fast and break things” is over — we’re now in the phase where scale collides with reality.
❓ FAQs
Q: What does Anthropic’s $900B valuation mean for everyday AI users? A: It signals that enterprise AI tools — especially coding assistants like Claude Code — are generating real revenue at unprecedented scale. It also means prices for high-performance AI usage will likely rise as companies chase profitability ahead of IPOs.
Q: Are AI data centres coming to NZ? A: Likely yes. The AI Blueprint for Aotearoa’s infrastructure workstream and the government’s focus on digital economy suggest data centre expansion is coming. The Australian backlash should serve as a cautionary tale about community engagement and environmental planning.
Q: Should I still learn to code if Google is using AI in interviews? A: Yes, but differently. The skill isn’t memorising syntax for a whiteboard — it’s understanding architecture, system design, and how to effectively direct AI tools. Google’s change validates that AI fluency is now a core engineering competency, not a cheat code.
🔍 THE BOTTOM LINE
The AI industry is entering its consolidation phase — trillion-dollar valuations, mass layoffs to fund infrastructure, regulatory recalibration, and an energy reckoning. The next 12 months will determine which companies survive the transition from “growth at all costs” to sustainable scale.
📰 SOURCES
- Financial Times — Anthropic $900B valuation report
- Reuters — OpenAI/Anthropic services acquisitions
- Business Insider — Google “Remy” personal agent
- The Decoder — Google/Meta personal agents race
- The Guardian — Australia data centre backlash
- NZ AI Forum — AI Blueprint for Aotearoa to 2030
- Newsroom — NZ AI return on investment MPs report
- Bloomberg — Meta layoffs details
- Euronews — EU AI Act simplification deal
- IEA — Energy demand from AI
- Bloom Energy — 2026 Data Center Power Report
- CIO — OpenAI/Anthropic enterprise push
- Quartz — Meta restructuring details
- The Verge — Google AI-assisted interviews