Eleven months after Brightmind Partners led SGNL’s Series A, CrowdStrike agreed in January to buy the identity security company for $740 million. That single trade is the argument the fund now carries into its second raise.
Brightmind closed a $250 million debut fund in late 2025 and has put it into 14 companies. One has sold. Stephen Ward, who started the New York firm in 2024 with Gur Talpaz, is now back in front of limited partners targeting roughly the same size for fund two, this time as sole general partner after Talpaz left in April with another early colleague.
The pitch rests on a view about timing that most venture firms state less plainly. Security startups sell best to strategic buyers between their third and seventh years, Ward says, and holding past that window means competing with the acquirer’s own roadmap.
Fund one’s exit fits the pattern.
SGNL had been in the portfolio less than a year when CrowdStrike moved. Some investors asked why Brightmind did not hold longer. Ward’s answer is that a sale into live demand beats defending a position past the point where a buyer’s build decision makes the asset redundant.
That is a harder discipline to run at scale than it sounds, because the money available to early-stage security companies has rarely been more plentiful. PitchBook counts $4.7 billion of seed-through-Series B cybersecurity deals through September, already past the whole of 2025. Deal count fell to 248 from 358 over the same comparison, so the dollars are concentrating in fewer, larger rounds. The average seed round in the sector rose to $8.4 million from $5.54 million a year earlier.
More capital per company changes what a fund has to do to earn its carry. A $250 million vehicle writing early checks needs outcomes large enough to return the fund, and the sell-at-year-five route caps the size of any single win. Brightmind’s answer is volume of judgement rather than volume of capital: a small team tracking which acquirers are building and which are buying, and pricing a company against that map instead of against a growth curve.
For limited partners the trade-off is legible. A fund that sells early converts paper gains into cash sooner and gives up the tail. Fewer than 20 of them backed the first vehicle, which suggests a concentrated register, not a broad one, and a second close at the same size would test how many of them accept the same bargain twice.
Ward’s read on buyer behaviour is specific enough to be testable. If one platform company is likely to move on a capability, the others follow, first out of competitive anxiety and then because their own customers ask. A target that overlaps with what the buyer already ships is a poor fit regardless of quality, and so is one whose adoption would force the buyer’s installed base to migrate off something it bought years ago. Those are the constraints he says the firm spends most of its time mapping.
The market he is raising into also has a definition problem. PitchBook tagged 278 deals in the first half of 2026 as AI security. Brightmind’s own review found 30 that qualified. The gap matters for a fund whose second vehicle will steer about 20% toward AI infrastructure, because a category that expands by relabelling makes comparable pricing difficult and makes diligence a matter of reading code and deployment architecture, not slide decks.
Where Ward draws the line is at the model layer. He does not expect the large model developers to absorb security, arguing they will supply enough capability to be useful and not enough to replace specialist tools. The companies he favours run detection on the customer’s own infrastructure, which keeps latency low and data inside the perimeter, a property that matters more as agents take actions rather than answer questions.
The fund’s own structure has changed since the first close, and not only at the top. Brightmind is five people. Iva Messy, who came from Felicis and Andreessen Horowitz, runs operations. Sven Wollschlaeger, who worked with Ward at Insight Partners, is a principal. Jamie Mneimneh, hired by Ward at Home Depot as a distinguished engineer, is CTO and built both the AWS environment where portfolio products get tested and the agents that now surface most of the firm’s deal flow. Taylor Cleveland, Ward’s former chief of staff, joined a few months ago.
Sourcing by software is a cost decision as much as a method. A firm with five people and a $250 million fund cannot carry an analyst bench, and management fees only stretch so far. Whether agent-driven sourcing produces the same quality of pattern recognition as a room of people arguing about a company is unresolved, and the second fund will be the test of it.
The split with Talpaz is part of the same story about a firm learning what it is. Both sides describe the separation in neutral terms. Talpaz and Tayler Sipperly, the firm’s third employee, have since started Cognition with CrowdStrike’s global CTO, reportedly targeting a $170 million first fund for the same sector, and they remain on the boards of several Brightmind companies. Two funds, one market, overlapping boards.
Ward’s route into this work did not run through finance. Four years of Coast Guard search and rescue, nine in the Secret Service, a stint running a flipped hacker who later accused him of being part of the scheme, roughly 18 months clearing his name, then security leadership at J.P. Morgan, TIAA and Home Depot. He was 49 when he started raising. Fewer than 20 limited partners filled fund one.
That biography explains the exit posture better than any market view. Someone who has watched a single relationship undo years of work tends to price the moment of sale, not the story around it.
What the second fund has to prove is narrower than the first. Fund one could be judged on whether the thesis produced one clean exit. Fund two has to show the thesis repeats while early-stage security capital keeps setting records and the number of genuine AI security companies stays roughly where Ward’s team counts it.