Nvidia Buys Hugging Face for Over 11 Billion Euros
Eva Mickler
4 min read Nvidia is acquiring Hugging Face for around 11.1 billion euros; the contract was signed on ...
Read Article
5 min read
On April 8, 2026, Gartner released a forecast that has direct budget implications for CIOs: the global semiconductor revenue will exceed the 1.3 trillion dollar mark for the first time in 2026 – a 64 percent growth compared to the previous year. The driver is well-known. The consequence for IT budgets is less so.
5 Min. Read Time
Key Takeaways
- 1.3 trillion dollars and 64 percent growth. Gartner forecasts the highest semiconductor growth in two decades for 2026. The drivers are AI accelerators for hyperscalers, whose infra spending grows by more than 50 percent.
- AI chips account for 30 percent of the market. AI semiconductors already make up a third of the total semiconductor revenue. The rest is distributed among memory, commodity chips, and non-AI applications.
- Memflation directly impacts non-AI budgets. DRAM prices rise by 125 percent in 2026, NAND flash by 234 percent. Server, storage, and device refresh: those who wait will buy at a higher price. Gartner expects price drops no earlier than late 2027.
- GPU capacity is contractually allocated. Hyperscalers secure available GPU production with multi-billion commitments. Enterprise customers compete with cloud infrastructure expansion for the same chips.
Why the Gartner figure is more than just market research for CIOs
A 1.3 trillion dollar market sounds like macro statistics. It’s not. Semiconductors are the supply chain behind every server, notebook, storage array, and GPU instance that CIOs purchase. What happens in this industry lands on the price lists of Dell, HPE, Lenovo, and in the cloud spot prices of hyperscalers with a 12 to 18 month delay.
The specific problem for 2026 and 2027: Gartner calls the phenomenon memflation. DRAM prices rise by 125 percent, NAND flash by 234 percent. Both are core components in almost every device. End-customer prices for hardware will follow – with a delay, but they will follow.
CIOs who have planned refresh cycles for server infrastructure or device fleets for 2026 should review their timing assumptions. Those who buy now will purchase before the price increase. Those who wait until 2027 will buy in a memory market without relief.
AI Chip Availability: The Underestimated Procurement Challenge
AI semiconductors account for 30 percent of the total semiconductor market – and they are scarce. AWS, Microsoft, Google, and Meta secure available GPU capacity for the next two to three quarters with multi-billion commitments. NVIDIA, AMD, and custom chip manufacturers prioritize these buyers.
For companies that need GPU capacity for their own AI workloads, two realistic scenarios emerge.
Cloud-first: Rent GPU capacity via AWS, Azure, or GCP instead of buying. No procurement lead time, no depreciation risks. But: cloud spot prices for A100 and H100 instances have also risen in 2026 because hyperscalers pass on the scarcity.
On-prem with long lead time: Those planning physical GPU servers currently expect 12 to 20 weeks of procurement time. That’s not a problem if the project starts in nine months. It’s a problem if the board has scheduled the AI initiative for Q3 2026.
Classification
64% semiconductor growth in 2026. Hyperscaler infra spending: more than +50%. AI chips: 30% of the total market.
Source: Gartner, April 8, 2026
Three Budget Adjustments for the Ongoing Planning Process
1. Calculate hardware refresh with a 15-25% buffer. DRAM and NAND memflation will drive up the cost of servers, storage, and end devices through 2026 and 2027. Those planning with last year’s price lists will need subsequent approvals.
2. Align AI workload timing with procurement reality. If an AI pilot needs hardware in Q3 2026, the order should be placed now. Lead times of 12 to 20 weeks for GPU hardware are standard in 2026 – not an exceptional state.
3. Realistically increase cloud budget for GPU capacity. Those replacing on-premises GPU with cloud spot capacity need a buffer. Gartner data shows a correlation between the hardware market and cloud pricing for AI instances.
The Gartner forecast doesn’t reveal new problems; it provides numbers for issues CIOs already know about. The value lies in underpinning board discussions about hardware budget adjustments with an external reference.
Frequently Asked Questions about the Gartner Semiconductor Forecast 2026
Does memflation affect cloud costs or only on-premises hardware?
Both, but with different timing. Cloud providers amortize hardware costs over several years and absorb some of the costs in the short term. In the medium term – Gartner expects more realistic memory prices by the end of 2027 at the earliest – higher costs will be factored into compute and storage tariffs. In the 2027 budget forecast, you should expect 10 to 20 percent price increases for memory-intensive cloud instances.
Are there alternatives to NVIDIA GPUs to avoid bottlenecks?
Yes, with limitations. AMD Instinct MI300X is available and competitive for inference workloads. For training on large models, the NVIDIA ecosystem (CUDA, toolchain) is still significantly more mature. Custom AI chips from cloud providers (AWS Trainium2, Google TPU v5) are interesting for specific workloads but tie you strongly to one provider. For initial enterprise AI projects: most workloads run on inference – alternatives to NVIDIA are more realistic than often assumed.
How accurate are Gartner semiconductor forecasts with such growth rates?
Historically reliable for the current year (deviations under 10 percent) because a large portion of capacity is already contractually bound. The 64 percent growth rate is exceptional but is supported by concrete order backlogs at NVIDIA, TSMC, and SK Hynix. The actual risk is not a downside risk but a supply risk: if TSMC capacities become scarcer than expected, prices will continue to rise.
Editor’s Picks
More from the MBF Media Network
- → Amazon Connect Agentic AI: Four New AI Solutions for SMEs (MyBusinessFuture)
- → OpenAI on Amazon Bedrock: What the AWS Launch Means for DACH Cloud Teams (cloudmagazin)
Source: Title image: Pexels / Brett Sayles (px:5050305)
Read more
- Technical Debt: Why the Board Must Act Now
- Agentic AI without an owner: Who is liable when the AI agent makes a mistake
- 725 billion US-Dollar CapEx: What the hyperscaler bet means for DACH-CIOs
Image source: AI-generated (Juni 2026), C2PA certificate embedded