The news
Three funding rounds in September 2026 show where enterprise spending on AI agents is going: Cognition raised more than $2 billion at $48 billion, Temporal $550 million at $12.55 billion, and Factory $200 million at $5 billion.
Cognition, which sells the Devin software-engineering agent, announced its round on September 8, led by new investors Andreessen Horowitz and Accel. The company said its revenue run rate rose from $492 million at its May round to almost $900 million; that May round valued it at $26 billion, Unite.AI reported. Named customers include Nvidia (NVDA), GE Aerospace (GE), Citi (C) and Mercedes-Benz (MBG.DE). New features let Devin investigate incidents, hunt for security vulnerabilities and start work automatically when events occur in Slack, GitHub or Linear, without anyone opening a chat.
Temporal, based in Bellevue, Washington, sells what it calls durable execution: software that saves the state of long-running processes so they recover from failures without being rebuilt. It announced its round on September 14, led by Lightspeed with Wellington Management, Goldman Sachs Alternatives and Tiger Global as co-leads. Temporal said annualized revenue grew more than 200% year over year, paying customers rose 139% to more than 4,300, including OpenAI, Snap (SNAP) and Nvidia, and it processed 1.9 trillion billable actions in August. Tech Funding News put its run rate above $250 million; its February round valued it at $5 billion.
Factory, whose coding agents are called Droids, announced its round on September 15 with investors including Blackstone (BX), Khosla Ventures, Sequoia Capital, Insight Partners and NEA. It named Nvidia, Blackstone, Royal Bank of Canada (RY), Palo Alto Networks (PANW), Adobe (ADBE) and T-Mobile (TMUS) as customers but did not disclose revenue. On September 18 it launched Factory Private, which runs its software inside a customer's own cloud account, data center or air-gapped network, and said a FedRAMP version for US government customers is in progress.
The common thread is operational control rather than raw model capability. Temporal chief executive Samar Abbas said AI has raised the cost of skipping reliability. Factory said on September 9 that enterprises can buy it through Anthropic's Claude Marketplace using existing spending commitments, and it markets a model router it says cuts aggregate costs 63% compared with frontier-model rates.
The numbers
- Cognition valuation
- $48 billion (up from $26 billion in May)
- Cognition revenue run rate
- Almost $900 million (from $492 million)
- Temporal valuation
- $12.55 billion (up from $5 billion in February)
- Temporal paying customers
- More than 4,300
- Temporal billable actions, August
- 1.9 trillion
- Factory valuation
- $5 billion
Why CEOs should care
For CIOs and CTOs, these rounds show agent budgets moving from chat assistants into engineering and operations workflows. Lightspeed partner Anoushka Vaswani described the problem in remarks quoted by Tech Funding News: “the demo is easy, production is hard.” Before scaling, ask vendors what happens when an agent fails midway through a multi-step task, where the agent runs and who holds its credentials, and how its output is measured. Factory's private and FedRAMP deployments exist to answer the second question for regulated buyers.
For CFOs, pricing is shifting toward usage. Temporal bills on actions it processes, and Factory pitches routing across models to cut inference costs. Finance teams should forecast agent spending as a variable cost tied to workload, set budget caps by team, and track spending against measurable output, such as incidents resolved or code merged, rather than license counts.
For CISOs, agents that start themselves from Slack or GitHub events and scan code for vulnerabilities are powerful and risky. Each agent needs its own identity, narrowly scoped permissions and an audit trail. Security teams should be part of agent procurement from the start, and should ask whether the vendor can run inside the company's own network.
The bigger picture
Investors are paying for momentum. Cognition's valuation rose from $26 billion to $48 billion between late May and early September, and Temporal's more than doubled since February. Crunchbase reported that AI-focused companies took more than 70% of global venture funding in the second quarter of 2026, and it counts SpaceX's $60 billion acquisition of Cursor maker Anysphere as the largest startup acquisition on record, which gives investors in AI coding tools a very large reference point for exit values.
Competition is intense at each layer. Temporal faces smaller orchestration rivals such as Inngest, Trigger.dev and Restate, Tech Funding News reported. Factory sells alongside the models it routes between, including through Anthropic's own marketplace, which means its value depends on the workflow and control it adds on top.
What’s next
Watch whether Cognition crosses $1 billion in run-rate revenue and discloses customer retention, whether Factory publishes revenue, and whether enterprises move from team pilots to company-wide agent contracts. If routing tools and falling model prices keep pushing inference costs down, agent vendors will need to prove value through reliability and workflow integration rather than model access alone.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
