One-Line Summary

OpenAI closes a record $122B round and plots an AI superapp, while Microsoft pushes multi-model Copilot and Anthropic formalizes an AI safety and data pact with Australia.

Big Tech

OpenAI raises $122B at an $852B valuation and teases an AI superapp

OpenAI closes a monumental funding round of $122 billion at an $852 billion post-money valuation, citing flywheel effects across consumers (900 million weekly active users and 50 million subscribers), enterprises (now 40% of revenue), developers (15 billion tokens per minute), and compute as its strategic advantage; it claims revenue has hit $2 billion per month, growing faster than the internet and mobile-era giants at similar stages. ¹

The round is anchored by Amazon, NVIDIA, and SoftBank, with continued Microsoft participation, plus a novel retail channel that contributes over $3 billion and inclusion in certain ARK Invest ETFs; OpenAI also expands its revolving credit facility to about $4.7 billion with a global bank syndicate, boosting flexibility to invest in infrastructure and scale. ^{1 2}

Strategically, OpenAI outlines a broadened multi-cloud and multi-silicon stack spanning Microsoft, Oracle, AWS, CoreWeave, and Google Cloud, plus NVIDIA, AMD, AWS Trainium, Cerebras, and its own Broadcom-partnered chip — underscoring compute as the compounding edge; separate coverage frames OpenAI’s use of Google Cloud for some ChatGPT workloads as a meaningful win for Alphabet’s cloud business. ^{1 3}

OpenAI says it is building a unified “AI superapp” that fuses ChatGPT, Codex, browsing, and agentic capabilities into one “agent-first” experience, turning consumer familiarity into the front door for enterprise usage; paired with GPT‑5.4 improvements and Codex’s 2 million weekly users (up 5x in three months), the company positions itself to capture more of the value from agent-driven workflows. ¹

Microsoft unveils multi-model Copilot and expands Cowork early access

Microsoft introduces a “Critique” workflow in Copilot’s Researcher agent where OpenAI’s GPT drafts and Anthropic’s Claude reviews responses, aiming to cut hallucinations and improve reliability; a “model Council” feature lets users compare model outputs side by side, moving Copilot from single-model to orchestration of best-of-breed systems. ⁴

The company broadens early access to Copilot Cowork via its Frontier program, leaning into agentic AI and responding to competition from Gemini and Claude Cowork; Microsoft frames the multi-model approach as both a productivity play and a trust signal for enterprises evaluating outcome quality. ⁵

Analysts read this as a shift from “model supremacy” to “model composition,” similar to choosing the right tool per job; if successful, it could make Copilot the default AI “router” inside Microsoft 365 workflows, reducing vendor lock-in fears and potentially speeding adoption across regulated teams. ⁶

Anthropic signs Australian federal MoU on AI safety, data, and infrastructure

Anthropic signs a memorandum of understanding with Australia to share Anthropic Economic Index data on AI adoption and labor impacts, participate in joint safety evaluations, and collaborate with universities; it also sets aside $3 million in Claude API credits for leading research institutes. ⁷

The MoU’s five pillars span tracking frontier AI progress and safety, productivity and workforce impacts, supply chain security, and building a vibrant domestic AI economy; the agreement mirrors safety cooperation in the US, UK, and Japan and aligns with Australia’s National AI Plan to attract data center investment and AI skills. ⁸

Local reporting highlights Anthropic’s intent to invest in renewable energy for data centers and the political optics of safety-focused partnerships, while experts note Australia must expand similar guardrails to other frontier labs to be effective; resource use, including high water consumption by large data centers, is flagged as a public concern. ^{9 10}

Industry & Biz

Depthfirst raises $80M (total $120M) at $580M valuation to build “general security intelligence”

AI-native cybersecurity startup Depthfirst secures $80 million Series B (total $120 million) at a $580 million valuation, aiming to develop domain-specific models that autonomously find and fix vulnerabilities; revenue is up 300% over two quarters with 20+ customers spanning multinationals and AI startups. ¹¹

The company launches its first in-house model, dfs‑mini1, trained with reinforcement learning to detect crypto smart contract flaws and evaluated on OpenAI’s EVMBench; Depthfirst claims it outperforms frontier models at 10–30x lower cost and can generalize to broader security tasks. ¹²

Investors liken its strategy to category winners that owned key threat vectors (e.g., Wiz in cloud security), arguing specialized AI, proprietary evaluation, and integration into developer workflows (with roughly 80% fix acceptance) can convert point features into a defensible product platform. ¹²

Community Pulse

Hacker News (10 points) — Mixed-to-negative: users criticize Microsoft’s Copilot bloat and UI clutter, citing awkward placements and ad-like surfaces.

"I hope Google follows their lead, Google Cloud has way to much space dedicated to "in platform ads"" — Hacker News

"tbh the fact that they put copilot in the snipping tool tells you everything about how those decisions were being made lol" — Hacker News

What This Means for You

OpenAI’s $122B raise signals a new normal: frontier AI now finances like national infrastructure. For teams, that means faster model upgrades, broader features (memory, personalization, multimodal), and likely falling per-token costs as compute scales — but expect pricing tiers that steer you into the coming “superapp.” ¹

Microsoft’s multi-model Copilot is a pragmatic hedge: quality beats loyalty. If your org hesitated over hallucinations, the GPT+Claude “Critique” flow and model comparisons make it easier to justify trials in research, marketing, and operations where accuracy is table stakes. Treat it like bringing two editors to the same draft. ^{4 6}

Regulation is getting hands-on. Anthropic’s MoU shows governments want real-time economic and capability telemetry from labs — expect procurement nudges for safety-evaluated systems and incentives tied to local infrastructure (and sustainability) commitments. If you sell into the public sector, prepare to show safety evals and usage impact data. ^{7 8}

Security teams should watch Depthfirst and peers: specialized, RL-trained agents that slot into dev workflows can shift AppSec from reactive tickets to proactive patches. If the 10–30x cost advantage and 80% merge rate hold up, AI-first security becomes a cost and speed lever, not just a shield. ^{12 11}

Action Items

1. Pilot Microsoft’s model Council in Copilot: Run a side-by-side bake-off (GPT vs. Claude) on one live workflow (e.g., research summary, RFP response) and document quality deltas and time saved.
2. Apply for Anthropic’s startup API credits: If you’re a VC-backed startup, apply for up to USD 50,000 in Claude API credits and prototype an internal agent for one ops task.
3. Book a Depthfirst demo for AppSec leads: Have dev and security teams test dfs‑mini1 on a scoped codebase (or smart contracts) and measure fix acceptance and MTTR impact.
4. Stress-test multi-cloud plans: Map one critical AI workload to a second provider (e.g., AWS→Google Cloud) to evaluate latency, egress costs, and failover — mirroring OpenAI’s diversified stack.