OpenAI’s February 2026: $110B Funding, Security Warnings, and Enterprise Deals
From a $110B funding round to new security work, defense deals, and deep partnerships with Amazon, Microsoft, and Snowflake, OpenAI just reset the AI roadmap for 2026.
February 2026 was a turning point month for OpenAI — and for anyone building or regulating AI.
In just a few weeks, the company announced a $110B raise, new work on disrupting malicious uses of AI, a high‑stakes agreement with the U.S. Department of War, and deeper integrations with enterprise giants like Amazon, Microsoft, and Snowflake
Key takeaways
- OpenAI raised $110B at a $730B valuation, with huge investments from SoftBank, NVIDIA, and Amazon, securing long‑term frontier compute capacity.
- A new “Disrupting malicious uses of AI” update details how attackers are weaponizing models and how OpenAI is trying to detect and disrupt abuse.
- Enterprise and infrastructure deals — including a multi‑year $200M Snowflake agreement, a long‑term Amazon partnership, and a renewed Microsoft alignment — pull OpenAI’s models closer to core enterprise and government workloads.
- February also brought defense, research, and safety moves: a Defense Department pact, new GPT‑5.x science and coding results, mental‑health guidance, and features like Lockdown Mode and Trusted Access for Cyber.
1. The $110B moment: scaling AI for everyone?
OpenAI’s “Scaling AI for everyone” announcement revealed $110B in new funding at about a $730B valuation, including a $50B Amazon commitment plus major checks from SoftBank and NVIDIA.
The raise locks in gigawatt‑scale GPU compute and new data centers to train and deploy GPT‑5‑class and successor models over several years.
For startups, that likely means renting top‑tier capability instead of training rival frontier models.
For enterprises, the question shifts from “Should we use OpenAI?” to “How do we safely wire these models into our data and governance stack?”
2. Security front: disrupting malicious uses of AI
OpenAI’s “Disrupting malicious uses of AI” update marks a move from policy to operational security.
It shows how attackers chain models with websites, messaging apps, and social platforms to scale phishing, influence operations, and fraud.
OpenAI details active interventions: detecting coordinated abuse, working with platforms, and throttling or removing malicious activity early.
For regulated sectors like healthcare, finance, and elections, frontier models now sit on both sides of the equation — the threat surface and the defense stack
3. Defense and geopolitics: the Department of War agreement
A new agreement with the U.S. Department of War brings OpenAI models into classified environments under “layered protections,” safety red lines, and legal constraints.
Reporting suggests the pact will place OpenAI systems on secure government networks after earlier friction with other vendors.
OpenAI stresses limits on autonomous targeting and oversight mechanisms, but the move raises hard questions about escalation, export controls, and AI in military decision‑making.
It cements frontier AI as part of hard‑power infrastructure, not just productivity tooling.
4. Enterprise gravity: Amazon, Microsoft, Snowflake, and the data stack
February also tightened OpenAI’s ties to major clouds and data platforms.
A long‑term Amazon partnership pairs a $50B investment with plans to bring OpenAI models closer to AWS customers on NVIDIA‑powered infrastructure.
A joint statement with Microsoft reaffirmed deep collaboration across research, engineering, and product, signaling a multi‑hyperscaler strategy rather than a single‑cloud bet.
Meanwhile, a $200M Snowflake deal brings GPT‑5.2‑class models into Snowflake Cortex AI and Snowflake Intelligence so customers can run OpenAI directly where their data lives.
For data leaders, that “bring the model to the data” pattern offers frontier models without losing control over where data sits or how it’s governed.
5. ChatGPT Frontier and the agent layer
In parallel, OpenAI introduced OpenAI Frontier, an enterprise platform for building AI agents that behave more like coworkers than chatbots.
Frontier gives agents shared context and permissions so they can run multi‑step workflows across CRMs, ticketing tools, and data warehouses — not just answer a single question.
To accelerate adoption, OpenAI announced Frontier Alliances with firms like McKinsey, BCG, Accenture, and Capgemini, positioning Frontier as the semantic layer on top of existing business software.
Combined with the Amazon, Microsoft, and Snowflake deals, Frontier shows that February’s story isn’t just bigger models, but a full agent platform wired into enterprise operations.
6. Research highlights: GPT‑5.x, coding, and scientific discovery
On the research side, OpenAI highlighted GPT‑5.2’s role in deriving a new theoretical physics result, framing large models as tools for original science.
It also updated how it evaluates coding, stepping back from SWE‑bench Verified as a frontier benchmark.
New releases GPT‑5.3‑Codex and GPT‑5.3‑Codex‑Spark aim at more powerful, agent‑style coding workflows, supported by fresh system cards.
The direction of travel is away from one‑shot completions toward runtimes where AI can iteratively read, write, test, and ship code.
7. Safety and product updates: Lockdown Mode, Trusted Access, and ads
Safety posts introduced Lockdown Mode and Elevated Risk labels in ChatGPT, plus Trusted Access for Cyber to handle security‑sensitive use cases under stricter controls.
An update on mental‑health‑related work refined guardrails and escalation paths when users seek help, clarifying what models should not attempt.
On the product side, experiments with ads in ChatGPT and higher‑throughput access to Codex and Sora, backed by new harnesses, app servers, and in‑house data agents, show how much infrastructure now supports an “agent‑first” model stack.
How February lands across the ecosystem
Investors see the $110B round as locking frontier‑scale compute into a few labs and pushing others toward niches, open‑source, or vertical plays.
Enterprise and data leaders read the Amazon, Microsoft, Snowflake, and Frontier moves as OpenAI moving into core cloud and data stacks, not sitting on the edge.
Security and policy teams view the malicious‑use update and Defense pact as proof that frontier models now sit inside both cyber and military workflows.
For startups and open‑source communities, the challenge is differentiation when the strongest models come from a few giants — via data, workflow, trust, or explicit bets on alternatives.
What’s Next?
Get more breakdowns like this in your inbox. Subscribe to The AI Entrepreneurs newsletter for weekly bite‑sized tutorials, tools, and playbooks to build smarter, faster, and with less guesswork. Join 70K+ founders and creators at AI Entrepreneurs — STANDOUT DIGITAL.
