The Exit Review

Published on : March 10, 2026

Yesterday, OpenAI announced it would acquire Promptfoo, an AI security startup founded in 2024 that helps enterprises red-team and evaluate LLM applications before shipping them. The deal terms were undisclosed. Promptfoo had raised $22.7 million across two rounds, carried a post-money valuation of roughly $85.5 million, and employed 11 people. The technology will fold into OpenAI Frontier, the company’s enterprise platform for building and operating AI agents. This is the latest in a string of OpenAI acquisitions in 2025 and 2026 that collectively signal a shift in what platform buyers want from early-stage companies. If you are building at the seed stage right now, particularly in AI infrastructure or developer tooling, this deal is worth studying. Not because the numbers are large, but because the profile of the company that got acquired looks a lot like the company you might be building right now.

TL;DR

• OpenAI is acquiring Promptfoo, an 11-person AI security startup, to integrate red-teaming and evaluation tools into its Frontier enterprise platform.
• This is OpenAI’s fifth known acquisition since January 2025, pushing its total disclosed M&A spend past $7.6 billion.
• Across 1,445 US VC-backed M&A transactions since January 2025, AI-tagged deals account for over $335 billion in disclosed value, and 89% of acquired companies were formerly VC-backed.
• AI security is now a distinct acquisition category with multiple platform buyers actively transacting.
• Promptfoo’s deal profile (small team, open-source adoption, Fortune 500 penetration, sub-$100M valuation, no disclosed revenue) fits a pattern early-stage founders need to understand: buyers will pay for distribution and technical trust, even without scaled revenue.

What Investors Look for at Seed Stage (and What This Buyer Actually Paid For)

Promptfoo is not a revenue-scale business. It is an adoption-scale one. The company built an open-source CLI and evaluation framework used by over 100,000 developers and more than 25% of the Fortune 500. That kind of penetration from an 11-person team is unusual, and it tells you exactly what the buyer was optimizing for: embedded developer trust in a category (agentic AI security) that has no incumbents yet. Founders preparing pre-seed decks or raising their first institutional round should look at the shape of this company at the time of acquisition and ask what it implies about how buyers evaluate early-stage startups.

• Promptfoo’s tools detect prompt injections, jailbreaks, data leaks, tool misuse, and out-of-policy agent behavior during development, not after deployment.
• The platform embeds red-team tests directly into CI/CD workflows, closing the loop between vulnerability discovery and remediation.
• OpenAI will integrate these capabilities natively into Frontier, its enterprise agent platform launched in February 2026.
• The open-source project will continue to support multiple LLM providers (including competitors), a signal that OpenAI wants Promptfoo to be a standard, not just an internal feature.
• Co-founders Ian Webster and Michael D’Angelo previously scaled AI systems at Discord to 200 million users. Buyer confidence in the team mattered as much as the product.

The playbook here is one founders should internalize: build the tool developers already trust, in a category where platform buyers have an urgent gap, and the exit finds you.

OpenAI Acquisitions in 2025 and 2026: A Buyer Map for Founders

Before Promptfoo, OpenAI had completed four acquisitions since January 2025: io Products ($6.5B), Statsig ($1.1B), Crossing Minds, and Jams. Five deals in fourteen months. Over $7.6 billion in disclosed spend. For a company that was, until recently, a research lab with no corporate development function, this pace is striking. Among the top acquirers across 1,445 US M&A transactions in our dataset, OpenAI now sits alongside Alphabet, Nvidia, and IBM. Each completed four or more acquisitions over the same period. Founders who are raising seed rounds for AI startups need to understand who these buyers are and what they are shopping for, because this buyer set increasingly defines which companies can exit and at what price.

• OpenAI’s targets span hardware (io Products), developer tooling (Statsig), recommendation systems (Crossing Minds), communication software (Jams), and now AI security (Promptfoo). The pattern is vertical integration across the full enterprise AI stack.
• If you are building something that touches AI agent deployment, evaluation, monitoring, or security, OpenAI is a plausible acquirer. That should inform how you build, how you position, and how you think about your competitive moat.
• The same multi-deal logic applies to other active buyers: Palo Alto Networks (3 acquisitions, $30B+ total), ServiceNow (3 deals), and Salesforce (2 deals, $8.8B combined). These are platform companies filling capability gaps through M&A, not making opportunistic bets.

For founders, the takeaway is tactical: know who is buying, know what they are missing, and build with that map in hand.

1,445 Deals, $621 Billion: The VC-Backed M&A Activity Founders Are Operating In

The dataset behind this analysis covers US M&A activity involving VC-backed companies from January 2025 through February 2026. It includes 1,445 completed transactions. Of the 321 deals with disclosed values, the aggregate totals approximately $621 billion. Software was the most active industry category at 652 deals, and AI-tagged transactions accounted for 439 deals worth over $335 billion combined. For founders planning their next twelve to eighteen months, the data says something simple: the exit market is active, and it has clear preferences. Early stage funding trends in the US point toward a market where the path from seed to acquisition is compressing, especially in AI-adjacent categories.

• 69 deals exceeded $1 billion. 145 fell between $100 million and $1 billion. 107 landed below $100 million. Both large platform consolidation and smaller strategic tuck-ins are running at high frequency.
• 89% of acquired companies (1,284 of 1,445) were formerly VC-backed. If you raised venture capital, you are operating in the primary pool that acquirers are fishing from.
• Monthly deal flow ranged from 76 to 132 transactions, with no sustained slowdowns across the fourteen-month window. The market did not pause.
• Life sciences and drug discovery dominated the value tables, but software and AI deals dominated volume. If you are building B2B software, the activity level is working in your favor.
• The top three deals by size: SpaceX/xAI ($250B), Synopsys/Ansys ($35B), Palo Alto Networks/CyberArk ($26.2B). Scale matters, but so does category timing.

The exit paths are real. The buyer set is concentrated. And the founders who understand both will have a meaningful edge when the conversation starts.

AI Security Startup Exits Are Forming a Category Founders Can Build Toward

Promptfoo’s acquisition does not exist in isolation. Across the same dataset, at least 78 deals involved cybersecurity-adjacent companies, and several of the largest specifically targeted AI security capabilities. If you are a founder building in this space, the signal is unambiguous: AI security has graduated from a niche concern to a category with multiple platform buyers willing to write nine- and ten-figure checks. That is rare at this stage of a market’s development, and it creates a real window for founders who move now.

• Palo Alto Networks acquired Protect AI for $700 million (AI/ML supply chain security). SentinelOne acquired Prompt Security for $275 million (generative AI prompt-layer security). F5 acquired CalypsoAI for $180 million (model-agnostic guardrails).
• CrowdStrike bought Pangea for $260 million, and Check Point bought Lakera for $300 million. None of these were acqui-hires. They were strategic transactions by major platforms racing to own the AI security stack.
• For founders at pre-seed or seed: if you are building tools that help enterprises test, monitor, or govern AI systems, the exit market for your category is maturing faster than most people expected.
• The buyer profile is consistent: platform companies consolidating a capability layer they believe will become table stakes. That means the window to build a defensible position is open now, but the pace of consolidation suggests it will not stay open indefinitely.
• This is also relevant for founders in adjacent categories like AI observability, compliance tooling, and agent monitoring. The buyers acquiring AI security startups are the same ones who will need those capabilities next.

If you are building best-in-class seed-stage B2B software in AI security, the question is no longer whether the market exists. The question is whether you can build enough distribution before someone else gets acquired into the slot you were targeting.

Acquisition Readiness at the Early Stage: What Promptfoo’s Deal Actually Tells You

Most founders do not think about acquisition readiness until a buyer shows up. That is a mistake, and Promptfoo’s deal illustrates why. The company had 11 employees, less than $23 million in total funding, a sub-$100M valuation, and no publicly disclosed revenue. What it did have: 100,000+ developers using its tools, penetration into a quarter of the Fortune 500, a category-defining open-source project, and a product that solved a problem every enterprise deploying AI agents will face. That is what made it acquirable. If you are building at the early stage, understanding what made this company a target is more useful than any term sheet template.

• Adoption without revenue is a viable exit profile when the buyer is acquiring a capability, not a business unit. If your product is embedded in developer workflows, that is a form of defensibility that revenue alone does not capture.
• Open-source distribution creates compounding trust. Proprietary startups at similar stages rarely achieve the kind of buyer-side confidence that comes from 100,000 developers choosing your tool voluntarily.
• Founding team credibility (Webster and D’Angelo scaled AI at Discord to 200M users) gave the acquirer confidence in execution. At the early stage, the team is often the primary asset being acquired. Your background is part of your positioning.
• The deal closed less than eight months after the Series A. That speed suggests the acquisition conversation started before or during the fundraise, which is common when a startup’s category attracts concentrated buyer interest.
• Founders should also watch for term sheet red flags during fundraising that could complicate a future exit: liquidation preferences that stack, anti-dilution provisions that shift economics, or board structures that slow deal approval. Acquisition readiness starts at the cap table.

The lesson is not that every 11-person startup can get acquired. The lesson is that founders who build distribution-first, in categories where platform buyers have urgent gaps, create acquisition optionality even at the earliest stages. Your pitch deck should reflect an understanding of who might buy you and why. Not because you are building to flip, but because that clarity sharpens every other decision you make.

The Buyer Map Is Readable Now, and Founders Who Use It Will Move Faster

One of the recurring patterns across these 1,445 deals is that the most active acquirers are not buying randomly. They are executing multi-deal strategies that map directly to their platform thesis. If you are raising a seed round or planning your next twelve months, understanding this buyer map is not optional. It defines which categories can exit, at what multiples, and on what timeline. For founders tracking early stage funding trends in the US, this data should be shaping your go-to-market and your narrative as much as your product roadmap.

• OpenAI (5 deals) is building a full-stack enterprise AI platform: hardware, developer tools, agent infrastructure, and now security. If your product fills a gap in that stack, you are on their radar whether you know it or not.
• Palo Alto Networks (3 deals, $30B+ total) is consolidating cybersecurity, observability, and AI safety under one vendor. Nvidia (4 deals) has moved beyond chips into AI infrastructure software.
• ServiceNow (3 deals) is layering AI-native capabilities onto its workflow platform. Salesforce ($8.8B across 2 deals) is acquiring data infrastructure and supply chain AI for its agentic roadmap.
• The companies getting acquired are the ones building for a known buyer’s strategic gap. Not the ones building in search of a use case. That distinction matters at every stage, from pre-seed deck to Series A positioning.
• For founders building in AI infrastructure, security, developer tooling, or enterprise automation: the buyer set is identifiable, the deal velocity is high, and the acquisition profiles are consistent. Use that information.
• Best VCs for product-led growth companies will tell you the same thing: the strongest early-stage companies are the ones that can articulate not just their market, but their exit map. That is not cynicism. It is clarity.

The exit market is not abstract. It has names, patterns, and preferences. The founders who study it will be better positioned than the ones who wait for a tap on the shoulder.

At Mighty Capital, we read product signals from a network of 600,000 product leaders to spot outliers before the market does. That means we recognize your value earlier, move with conviction, and back you on terms that keep you in control. Then we go further: we open our entire commercial platform to you, with a commitment to deliver $10 of value for every $1 we invest. We’ve already delivered $1B to our portfolio. Read more about our approach here: https://productsthatcount.com/?filter_tag=research