The data center arms race isn't cooling off. If anything, it's shifting into higher gear. AI servers are becoming the backbone of modern computing infrastructure, and Nvidia Corp (NVDA) is orchestrating the tempo while hyperscalers race to keep pace. Broadcom Inc (AVGO) and Advanced Micro Devices Inc (AMD) may not be steering this ship, but they're absolutely along for the ride.
According to Beth Kindig, CEO and Chief Investment Strategist at I/O Fund, Nvidia's GB300 is becoming the "main driver" behind the next wave of AI infrastructure investment. We're not talking about a cyclical uptick anymore. This looks more like a fundamental shift in how computing gets done.
The Numbers Behind the 2026 Buildout
TrendForce projects AI server shipments will climb more than 28% in 2026, picking up speed from roughly 24% growth last year. That acceleration tells you everything: the market isn't taking a breather after the initial AI spending wave. It's pushing harder.
Meanwhile, overall global server shipments? Expected to grow about 12.8%. The gap between those numbers is the story. Traditional computing infrastructure is chugging along steadily. AI compute is running full sprint. The buildout is getting deeper, more complex, and increasingly specialized.
GPUs are forecast to account for around 70% of AI server shipments, with Nvidia's GB300 positioned as the workhorse powering this expansion. But this isn't a solo act.
AMD and Broadcom Stake Their Claims
AMD's VR200 platforms are expected to pick up momentum in the second half of this year as cloud providers look for more options on inference-heavy workloads. That's where the real volume happens once models are trained and deployed.
Broadcom, meanwhile, is positioned to capitalize as hyperscalers demand faster interconnects, custom silicon designs, and high-speed networking gear to wire together massive AI clusters. When you're building infrastructure at this scale, the plumbing matters as much as the processors.
The Custom Chip Play from Google and Meta
There's a quieter trend developing alongside the GPU surge. ASIC-based AI servers are projected to hit nearly 28% of total shipments this year — the highest proportion in years. That's being driven by Alphabet Inc (GOOGL) (GOOG) Google and Meta Platforms Inc (META) scaling up their own custom chips for inference workloads.
This isn't an immediate threat to Nvidia's dominance, but it signals how the biggest cloud players are hedging their bets: lean on Nvidia for training the models, then shift to in-house silicon for the day-to-day inference work that runs at scale.
What 2026 Looks Like for Investors
Nvidia is lighting the fire. Hyperscalers are dumping fuel on it. Broadcom and AMD are riding shotgun, capturing their piece of the spending boom.
If this trajectory holds, 2026 is shaping up to be one of the most capital-intensive years in data center history. This isn't just another capex cycle. It's a fundamental rewiring of the infrastructure that powers the internet, driven by AI workloads that aren't slowing down anytime soon.
