June 12, 2026 – Silicon Valley’s “AI rollup” model jumps from quiet private deals to two large public-to-private takeovers.
In Summary
Venture firms now buy whole companies and rebuild them around AI.
Trian and General Catalyst took Janus Henderson private for about $7.4B.
Long Lake agreed to buy Amex GBT for about $6.3B in May 2026.
The Amex GBT offer carried a steep 60.2% premium to shareholders.
Venture capital has found a fresh way to grow. Instead of selling software to companies, top firms now buy whole businesses. They then rebuild each firm around artificial intelligence. Silicon Valley calls this strategy the AI rollup. Moreover, the model has now crossed from private deals into public markets.
What the AI rollup actually means
The AI rollup borrows its logic from software economics. General Catalyst managing director Madhu Namburi calls it “service as software.” The phrase reworks the familiar “software as a service,” or SaaS, label. SaaS firms grew profits because revenue rose faster than costs. Therefore, AI rollups try to replicate that math within service businesses.
Buyers acquire a legacy company first. Next, they add AI to cut manual, repetitive work. As a result, margins can widen without large new headcount. Venture firms have run this play since 2023, mostly in private. However, two large public deals have now pushed it onto Wall Street.
Janus Henderson goes private
The first big crossover involved a household asset manager. In December 2025, Trian Fund Management and General Catalyst agreed to buy Janus Henderson. The all-cash deal valued the firm near $7.4 billion. Shareholders were offered $49 a share. That price marked an 18% premium to the stock’s October 24 close.
Trian already held about 20.6% of the company. Furthermore, the buyers later raised the offer to $52 a share. Investors then approved the take-private in April 2026. Janus Henderson managed roughly $493 billion in assets at year-end. Chief executive Ali Dibadj framed the deal as support for the firm’s long-term plan.
Amex GBT follows at a steep premium
A second deal landed in May 2026. Long Lake Management agreed to buy American Express Global Business Travel. General Catalyst and Alpha Wave backed the bid. The all-cash transaction reached about $6.3 billion. Shareholders will receive $9.50 a share.
That offer sits 60.2% above the May 1 closing price. It also beats the 30-day average price by roughly 65%. Long Lake itself only launched in 2023. Notably, this marked the firm’s first public take-private. Big holders representing 69% of shares already back the deal. Amex GBT chief Paul Abbott pointed to AI-driven change in travel.
Why venture firms are on offense
The strategy reflects a shift in power across the finance sector. Traditional private equity bought enterprise software at peak prices last cycle. Those buyers now sit on defense. In contrast, venture firms see cheaper, AI-ready targets in older industries. PitchBook describes the model as a “Future of Services” bet.
Analysts note that the approach also helps firms sidestep normal venture limits. General Catalyst has co-created about a dozen rollup vehicles. Meanwhile, Joshua Kushner’s Thrive Capital runs Thrive Holdings on the same idea. That vehicle holds more than $1 billion in capital.
A timeline from private bet to public deal
The path to Wall Street took only a few years. Venture firms began quietly buying service businesses in 2023. The Janus Henderson deal then came to light in December 2025. Shareholders approved that transaction by April 2026. Soon after, the Amex GBT deal arrived in May 2026.
What to watch next
Several questions now hang over the AI rollup boom. First, buyers must prove that AI truly lifts margins. Second, regulators will review these large take-private deals. Third, traditional private equity may copy the approach. Still, the early momentum already looks clear.
Venture capital has moved from selling tools to owning the customer. For investors, the next real test is execution, not ambition. Consequently, the results inside these private firms will matter most.
