Microsoft’s Delayed AI Chip Gives Nvidia Even More Breathing Room

Setbacks push Microsoft’s Maia “Braga” chip to 2026, shaking confidence in its silicon ambitions

Microsoft’s big bet on building its own artificial intelligence chips has hit another wall. Its next-generation Maia processor—nicknamed “Braga”—won’t hit mass production until 2026. That’s at least half a year late, and that lag is turning into opportunity for the company it hoped to outmaneuver: Nvidia.

Microsoft started developing its custom chips to loosen its dependency on Nvidia’s pricey GPUs, which currently dominate the AI infrastructure market. But sources now say the Braga chip has run into repeated delays—staffing shortages, redesigns, and high turnover are all getting in the way. And with Nvidia’s Blackwell lineup on track, the delay could leave Microsoft eating dust.

Delay gives Nvidia a wider lead in AI hardware race

Microsoft didn’t just build Braga for show. The chip was part of a high-stakes push to own the hardware powering the AI boom—especially for Azure, its cloud platform. The thinking was simple: custom chips = lower costs = more control.

But now, things look shaky.

A report from The Information says Braga is delayed by at least six months, pushing its launch window into early or even mid-2026. That’s a full year behind Nvidia’s new Blackwell GPUs, which are already generating buzz for their speed and power efficiency.

One sentence sums up what that means: Microsoft’s next chip may already be outclassed before it even ships.

Internal struggles weigh down chip team

Braga wasn’t supposed to be a moonshot. It was supposed to be a leap forward from the Maia 100, Microsoft’s first AI chip unveiled late last year. But people inside the company paint a different picture. The project has been bogged down by…

• Unplanned design changes, forcing reworks and rerouting

• Key departures from the silicon team, weakening continuity

• Difficulty hiring top chip engineers in a brutally competitive market

One former engineer called it “a perfect storm of mistakes.” Another said: “Everyone knew the schedule was ambitious, but few expected it to fall this far off.”

Azure’s AI plans hit turbulence

This isn’t just about silicon—Microsoft’s entire cloud AI vision is tied to Braga.

The company has spent billions scaling Azure into an AI-first cloud, powering everything from OpenAI’s ChatGPT to enterprise LLMs. Running those workloads on custom chips was supposed to cut costs and free Microsoft from Nvidia’s grip. It’s not panning out that way—at least not yet.

Executives internally are said to be growing impatient. One person briefed on the matter told TechSpot that there’s increasing pressure to show results by early 2026 or “risk losing serious credibility with enterprise customers.”

And there’s money on the line, big money. Microsoft reportedly spends billions annually on Nvidia GPUs. Delays in Braga mean continuing those purchases longer than planned—and paying premium prices along the way.

Nvidia just keeps winning

The longer Microsoft takes, the stronger Nvidia gets.

Already, Nvidia commands over 80% of the AI GPU market. Its new Blackwell chips—unveiled in March 2025—promise huge improvements in speed, memory, and energy use. Early benchmarks show performance leaps of 30% or more over the prior generation.

Meanwhile, Nvidia isn’t just selling chips. It’s bundling software, networking, and full-stack solutions, making it even harder for competitors to catch up.

Let’s look at what Microsoft is up against:

So yeah—Microsoft has a hill to climb, and it’s getting steeper by the month.

Silicon independence proves harder than expected

Big Tech wants to own its own chips. Apple did it with iPhones. Google’s doing it with Pixel and TPUs. Amazon has Trainium. Microsoft? Still stuck in neutral.

And making AI chips isn’t like designing a smartphone SoC. We’re talking bleeding-edge transistors, ultra-fast interconnects, heat management, software stacks, and hardware-software co-design—all at once. Not to mention: competition for talent is cutthroat.

The chip industry is not forgiving.

One analyst from SemiAnalysis noted: “It’s like trying to land a rocket on a dime—while the dime is moving.” Not impossible, but definitely harder than Microsoft’s early timelines suggested.

Braga delay may open door to other partnerships

Microsoft isn’t throwing in the towel, but it is hedging.

While waiting on Braga, the company has ramped up purchases of Nvidia and AMD hardware. It’s also reportedly exploring tighter partnerships with chip startups like Tenstorrent and Groq.

Meanwhile, internal chatter suggests Microsoft may consider dual-sourcing parts of its future AI workloads—splitting them between Azure-native silicon and third-party hardware, depending on performance and timing.

That hybrid model could cushion the blow. But it also underlines the fact that Microsoft isn’t ready—at least not yet—to go full solo in the AI chip race.