VC
Value Add VC
⚡HomePulse⚡Helpful Apps📝Blog
← Value Add PulseBIG TECH~$3.9B

Qualcomm to Buy AI-Software Startup Modular for ~$3.9B to Attack Nvidia's CUDA Moat

Qualcomm agreed to acquire Modular -- the startup behind the Mojo programming language and MAX AI platform -- in an all-stock deal worth roughly $3.9 billion. The bet is that the way to break Nvidia's dominance isn't a faster chip but an open software layer that lets AI models run efficiently across any hardware, directly challenging the CUDA lock-in.

~$3.9B
Deal Value
All-stock (~19.2M shares)
Structure
Modular (Mojo + MAX)
Target
~150 employees
Team
Chris Lattner, Tim Davis
Founders
TC
Trace Cohen
Early-stage VC & angel · Founder, New York Venture Partners
June 24, 2026
2 min read
KEY TAKEAWAYS FOR VCs & FOUNDERS
1

Nvidia's real moat is CUDA software, not silicon -- Modular is a frontal assault on exactly that

2

Qualcomm gets Modular's ~150-person team, including legendary compiler creator Chris Lattner

3

Hardware-agnostic AI tooling could let AMD, Intel and Qualcomm chips finally compete on software parity

4

A ~$3.9B all-stock price signals how much incumbents will pay to escape Nvidia dependence

TC
The VC Read · Trace's TakeTrace Cohen

Everyone obsesses over Nvidia's chips and misses that the moat was always CUDA -- the software developers can't quit -- so Qualcomm buying Modular is the smartest anti-Nvidia move of the year. The Chris Lattner factor alone may justify the price; he's built three foundational toolchains and that caliber of compiler talent doesn't come up for sale. The catch is the paradox of neutral platforms: the value of Mojo is that it's hardware-agnostic, and the moment a chipmaker owns it, the rest of the industry gets nervous. Watch whether the developer community stays -- if it does, this is how the CUDA monopoly finally cracks.

⚡ AI Chip Wars →

Qualcomm has agreed to acquire AI-software startup Modular in an all-stock deal worth approximately $3.9 billion, issuing up to 19.2 million shares, according to Reuters and Bloomberg. The transaction is expected to close in the second half of 2026 pending regulatory and shareholder approvals.

Modular built the Mojo programming language and the MAX AI platform -- software designed to make AI models run efficiently across heterogeneous hardware, including chips from Nvidia, AMD, Intel and others. In other words, Modular's entire thesis is to abstract away the dependence on any single vendor's stack, which makes it a direct strike at CUDA, the proprietary software layer that is the true source of Nvidia's dominance.

The deal is as much about talent as technology. Qualcomm inherits Modular's roughly 150 employees, including both founders: Tim Davis and Chris Lattner, the compiler luminary who created LLVM, Clang and Apple's Swift language. That kind of low-level systems talent is among the scarcest and most valuable in the industry, and acquiring it intact is arguably worth the price on its own.

“That kind of low-level systems talent is among the scarcest and most valuable in the industry, and acquiring it intact is arguably worth the price on its own.”

The strategic context is a coordinated assault on Nvidia. The same week, OpenAI unveiled its own custom inference chip with Broadcom, and the through-line is unmistakable: the industry has concluded that Nvidia's lead must be attacked on two fronts -- bespoke silicon for the biggest buyers, and open software to neutralize CUDA for everyone else. For Qualcomm, which has struggled to break into data-center AI against Nvidia and AMD, Modular is a credible wedge: give developers a hardware-agnostic toolchain, and Qualcomm's own accelerators suddenly have a path to adoption.

The numbers fit a pattern of premium prices for AI software leverage. A ~$3.9B all-stock deal for a company with a developer following but modest revenue is underwriting strategic position, not current cash flow -- the same logic behind AMD's ZT Systems purchase and the talent-driven 'acqui-hires' that have defined this cycle. The risk is integration: open, neutral software platforms can lose their cross-vendor credibility the moment a single chipmaker owns them, and Qualcomm will have to keep Mojo genuinely hardware-agnostic to retain the developer trust that makes it valuable.

What to watch: whether Modular's open-source community stays loyal under Qualcomm ownership, whether Mojo/MAX actually narrows the CUDA gap on AMD and Qualcomm silicon, and how Nvidia responds to a well-funded challenger aimed squarely at its least-defensible flank.

ShareXLinkedInEmail
More onNvidia →

Originally reported by Bloomberg. Analysis and editorial commentary by Value Add Pulse.

← Back to Pulse

Markets Now

live
SPCX▲+0.31%
$227.10
CBRS▼-20.20%
$256.80
SPY▲+0.08%
5,942.70
QQQ▲+0.11%
20,012.30
NVDA▼-1.42%
$152.90
MSFT▲+0.25%
$479.80
GOOGL▲+0.34%
$209.10
META▲+0.34%
$651.40

Read Next

BIG TECH~-20% stock drop

Cerebras Plunges ~20% in Its First Earnings as a Public Company on a Margin Scare

Cerebras shares fell roughly 20% -- back near their IPO price -- after the AI-chip maker's first earnings report as a public company spooked investors with a soft gross-margin outlook, even as it beat on revenue. CEO Andrew Feldman said the market 'misunderstood' the guidance, which reflects a temporary plan to lease back systems while it builds its own data centers.

BIG TECHNew product push

Mark Zuckerberg Wants Meta to Launch Its Own Prediction Market

Mark Zuckerberg is pushing Meta to build its own prediction market, according to TechCrunch, betting that the booming category -- popularized by Polymarket and Kalshi -- belongs inside Meta's social and advertising machine. The move would thrust Meta into a fast-growing but heavily scrutinized corner of online finance.

BIG TECH21,000 jobs cut

Oracle Cuts 21,000 Jobs to Help Fund Its Debt-Fueled AI Data Center Buildout

Oracle has laid off roughly 21,000 workers as it pours borrowed money into a massive AI data center expansion, according to Ars Technica. The juxtaposition -- deep headcount cuts alongside an aggressive, debt-financed compute buildout -- captures how far legacy software firms are stretching to stay relevant in the AI infrastructure race.

@Trace_Cohen·t@nyvp.com