AI data-center bottleneck drives $1B Seligman fund, $10.7B chip bets

AI capabilities are increasingly gated by physical infrastructure, and Seligman's $1 billion vehicle is betting that the next wave of large tech companies will be built in the data center stack. The firm is targeting accelerators, networking, power, cooling and cybersecurity—the constrained layers where software improvements alone cannot solve throughput and efficiency problems. With SambaNova challenging Nvidia in accelerators and Lumilens addressing optical interconnect, Seligman's portfolio maps directly to AI compute bottlenecks.

Seligman Investments has doubled the deployable capital of its venture arm to $1 billion less than a year after launching it, a striking escalation of its conviction that physical bottlenecks in the artificial intelligence buildout will mint the next generation of large technology companies. The announcement came on September 28, 2026, when executives told Reuters that Seligman Ventures, the private investment arm of the Silicon Valley-based technology investor, had expanded from the $500 million it launched with in February 2026 and had already put more than $300 million to work across 14 investments in AI hardware, connectivity and cybersecurity. The portfolio spans critical pieces of the data center stack, from SambaNova, a challenger to Nvidia in AI accelerators, to Lumilens, an optical component maker.

“

and Canadian startups hit a record $392 billion in the first half of 2026, with about $10.7 billion of that flowing into semiconductor startups on a pace to exceed last year's total.

The expansion marks a meaningful reversal of two decades in which Silicon Valley returns were built largely on software. Seligman's hardware-first approach reflects a broader market shift confirmed by Crunchbase data: venture investment in U.S. and Canadian startups hit a record $392 billion in the first half of 2026, with about $10.7 billion of that flowing into semiconductor startups on a pace to exceed last year's total. For an industry that spent years avoiding capital-intensive bets, the numbers suggest that AI compute demand has re-legitimized hardware as a venture-scale opportunity.

Seligman Ventures is not simply deploying capital; it is taking an unusually active position for a vehicle backed by a public-market manager. Unlike many crossover funds that buy into private companies as passive allocators, Seligman holds six board seats and three board observer seats. That operational involvement, paired with a barbell strategy that combines early-stage bets and late-stage or pre-IPO checks, is designed to capture returns across the maturity spectrum while maintaining influence over governance and execution. Managing partner Umesh Padval said the deal flow has been 10 or 15 times greater than he anticipated, and the firm intends to keep pursuing opportunities in accelerators, networking, power and cooling, as well as cybersecurity.

The move is also part of a broader structural change in private markets: public-market managers are reaching earlier into technology companies because startups are staying private longer. For Seligman, the $1 billion vehicle is a second attempt at venture investing for Paul Wick, chief investment officer of Columbia, Seligman's technology business, and the firm is clearly positioning itself to ride what it sees as a multi-year buildout of AI infrastructure rather than a short-term cycle.