---
title: "SemiAnalysis Significantly Downgrades AI Compute Demand Forecast: 2028 Capex Estimate Cut by Nearly 30%, Five-Year Wafer Shipment Area Reduced by 17%, NVIDIA Hit Hardest"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/305662589.md"
description: "SemiAnalysis has lowered its global AI accelerator capital expenditure forecast by 28% for 2028 and by 40% for 2030. The cumulative forecast for accelerator wafer shipment area from 2026 to 2030 has been reduced by 17%, while the HBM demand forecast for 2028 was slashed by 30%, and the power consumption demand forecast for 2030 was cut by 34%. NVIDIA emerges as the most impacted company, accounting for 66% of the global reduction in wafer shipments, with its 2030 shipment forecast based on compute revenue downgraded by as much as 66%"
datetime: "2026-11-16T11:31:58.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/305662589.md)
  - [en](https://longbridge.com/en/news/305662589.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/305662589.md)
generator: "portal-rs"
---

# SemiAnalysis Significantly Downgrades AI Compute Demand Forecast: 2028 Capex Estimate Cut by Nearly 30%, Five-Year Wafer Shipment Area Reduced by 17%, NVIDIA Hit Hardest

New signs of cooling have emerged in the AI infrastructure investment boom.

In its latest model update released on November 13, semiconductor research firm SemiAnalysis **significantly lowered its forecasts for AI accelerator shipments, capital expenditures, HBM demand, and power requirements for the coming years.** The adjustments are particularly pronounced for the period from 2028 to 2030: **the global AI accelerator capital expenditure forecast for 2028 was cut by 28%, followed by reductions of 41% and 40% for 2029 and 2030, respectively.**

According to the latest model, global AI accelerator capital expenditures in 2030 will drop from the previously projected $1.531 trillion to $923 billion, a decrease of $608 billion. Meanwhile, **the cumulative forecast for accelerator logic wafer shipment area from 2026 to 2030 has been reduced by 17%, with NVIDIA accounting for 66% of this reduction.**

**NVIDIA is the hardest hit in this round of forecast adjustments.** **SemiAnalysis not only significantly lowered its shipment forecasts for NVIDIA in the coming years but also further reduced the average selling price (ASP) assumptions for new-generation products, arguing that as competition in the AI hardware market intensifies, NVIDIA will need to accept lower margins to maintain competitiveness.**

## **AI Capital Expenditure Expectations Plummet, 2030 Forecast Cut by $608 Billion**

The most intuitive signal from SemiAnalysis's adjustment is the reassessment of the scale of global AI accelerator capital expenditures.

According to its latest model, the global AI accelerator capital expenditure forecast for 2026 remains largely unchanged at around $392 billion. However, starting in 2027, the previous high-growth expectations have been significantly revised downward.

**The capital expenditure forecast for 2027 was lowered from $770 billion to $627 billion, a 19% decrease; for 2028, it dropped from $1.054 trillion to $758 billion, a 28% decrease; for 2029, it fell from $1.352 trillion to $796 billion, a 41% decrease; and for 2030, it declined from $1.531 trillion to $923 billion, a 40% decrease.**

This means that the path of continuous high-speed expansion in AI accelerator capital expenditures previously expected by SemiAnalysis is being clearly corrected. Especially in 2029 and 2030, the annual investment scale, originally expected to exceed $1 trillion, has now been lowered to the range of $800 billion to $900 billion.

This does not mean that AI infrastructure investment will stop growing. According to the new forecast, global AI accelerator capital expenditures will still grow from $392 billion in 2026 to $923 billion in 2030, albeit at a scale significantly lower than previously anticipated.

Behind the capex downgrade are dual changes in AI industry revenue growth and the financing environment.

SemiAnalysis pointed out that the slowdown in the growth rate of new Annual Recurring Revenue (ARR) at frontier AI labs in the third quarter prompted a re-examination of the token revenue scale the entire ecosystem can generate. To ensure AI business revenue can more sustainably cover computing costs, the firm accordingly lowered its expectations for the expansion of future compute supply.

At the same time, debt and equity investors' willingness to invest in AI infrastructure capital expenditures is weakening, especially for projects lacking procurement commitments from large cloud service providers (Hyperscalers). This will directly impact the development of data center land, power, and building shells, leading to a reduction in the power supply available for deploying AI computing systems in the future.

The results of the U.S. midterm elections have further exacerbated this pressure. SemiAnalysis expects that **5.8 GW of U.S. data center supply originally planned to come online from the second half of 2027 to the first half of 2028 will be delayed, and regulatory factors will further compress data center capacity in 2028.**

However, the firm emphasized that the delay in data center construction caused by the midterm elections is only an additional factor in the forecast downgrade; the slowing growth of AI revenue and the tightening financing environment are the more important drivers.

## **Cumulative Wafer Shipment Area Down 17%, HBM Demand Also Significantly Revised**

Beyond the capex forecast downgrade, SemiAnalysis's judgment on actual hardware demand has also become more cautious.

According to the latest model, **the cumulative forecast for global AI accelerator logic wafer shipment area from 2026 to 2030 has been reduced by 17%. Specifically, the forecasts for 2027, 2028, 2029, and 2030 were lowered by 13%, 16%, 24%, and 15%, respectively.**

Notably, **NVIDIA accounts for 66% of the reduction in wafer shipment area, meaning that this adjustment in AI hardware demand expectations is not evenly distributed among suppliers; NVIDIA is the primary affected party.**

HBM demand forecasts have also seen significant downward revisions. The global HBM shipment forecast for 2028 was lowered from 9.0 EB to 6.4 EB, a 30% decrease; for 2029, it dropped from 10.3 EB to 7.9 EB, a 23% decrease; and for 2030, it fell from 10.9 EB to 9.6 EB, a 12% decrease.

SemiAnalysis believes that **the supply-demand balance for HBM is improving, leading to a downgrade in future HBM price assumptions, with these price changes passed through to the Bill of Materials (BOM) costs for various AI chips.**

Previously, a significant portion of the growth in AI infrastructure spending in 2027 came from price increases, particularly in memory prices. In the firm's view, this downgrade in price forecasts also indicates that the supply chain is returning to a more normal state.

Additionally, power consumption requirements for AI accelerators have been significantly downgraded. According to the latest forecast, the power demand for AI accelerators in 2030 will drop from 122.8 GW to 80.7 GW, a 34% decrease; the 2029 forecast was lowered from 108.7 GW to 72.5 GW, a 33% decrease.

This change stems from both reduced chip shipments and adjustments to assumptions regarding single-chip power consumption in the future. SemiAnalysis noted that electricity continues to be a limiting factor, and the maximum chip power consumption in future accelerator product roadmaps may also trend toward more moderate levels.

## **NVIDIA Under Greatest Pressure, Future Chip Shipment Forecasts Significantly Downgraded**

Among major AI chip suppliers, NVIDIA is the company most affected by this forecast adjustment.

SemiAnalysis **lowered its shipment forecasts for NVIDIA from 2027 to 2030 by 22%, 39%, 56%, and 66%, respectively, based on Compute Revenue**; calculated based on total power consumption including the entire system, the reductions were 14%, 35%, 50%, and 59%, respectively.

Regarding pricing, SemiAnalysis not only considered the impact of declining HBM costs but also argued that NVIDIA needs to reduce its markup on Feynman and subsequent products to cope with increasingly fierce competition.

Specifically, the ASP forecast for Feynman was lowered from approximately $110,000 to $78,000, Feynman Ultra from about $175,000 to $98,000, and Feynman Next from roughly $191,000 to $118,000.

**The simultaneous downgrade in both shipment volumes and selling prices will put greater pressure on NVIDIA's future revenue forecasts.**

SemiAnalysis explained that NVIDIA is more heavily impacted primarily because its chips have become less attractive compared to other platforms in the compute procurement portfolios of some large AI labs.

Taking Anthropic as an example, its four major computing platforms include TPUs, Trainium, AMD GPUs, and NVIDIA GPUs. SemiAnalysis believes that in terms of performance-to-cost ratio, NVIDIA GPUs are on the margin among these four platforms, leading Anthropic to prefer other options.

Regarding OpenAI, although GPUs still hold a significant position in its infrastructure, AMD GPUs have become an attractive alternative. Furthermore, OpenAI will prioritize the deployment of its in-house ASIC projects, such as Jalapeno and the next-generation Serrano.

Neocloud providers constitute another risk for NVIDIA. SemiAnalysis pointed out that NVIDIA has the highest exposure to Neocloud customers, especially smaller long-tail clients. These companies currently face greater project execution and financing risks.

Notably, SemiAnalysis emphasized that **this model adjustment did not incorporate the potential impact of Jensen Huang stepping down as NVIDIA CEO; even if Huang remains CEO, the firm would make the same forecast downgrades.** However, the firm believes that **leadership changes could further increase the downside risk of NVIDIA losing market share.**

## **Broadcom Relatively Stable, AMD and MediaTek Face Varying Degrees of Adjustment**

In sharp contrast to NVIDIA, Broadcom demonstrated greater resilience in this forecast adjustment.

SemiAnalysis lowered its cumulative revenue forecast for Broadcom's ASIC business from 2026 to 2030 by 11%, mainly due to ASP adjustments resulting from lower HBM costs, rather than large-scale cuts in shipment volumes.

The firm believes that **Broadcom's major ASIC projects, including Google's TPU and those related to OpenAI and Anthropic, have advantages in performance-to-cost ratio and energy efficiency. Additionally, Broadcom can provide financing support for data center land, power, building shells, and system deployment by collaborating with creditors.**

AMD also showed some resilience but still faces significant downgrades. Based on total power consumption, SemiAnalysis lowered its forecasts for AMD in 2028, 2029, and 2030 by 23%, 55%, and 57%, respectively; based on expenditure, the reductions were 29%, 35%, and 32%, respectively. The 2027 forecast remained relatively stable, with power consumption figures basically flat and expenditure down by 11%.

**AMD was less impacted than NVIDIA, mainly benefiting from its layout among large customers like Meta, OpenAI, and Anthropic.** SemiAnalysis believes that the warrant transaction arrangements between Meta and OpenAI with AMD make purchasing AMD systems highly attractive; Anthropic also views AMD as the second-best choice in terms of performance-to-cost ratio for inference computing, after TPUs.

Furthermore, AMD's recent product deployment schedule has actually accelerated. SemiAnalysis moved the shipment timeline for approximately 25,000 MI400 chips and about 30,000 Meta-customized MI450-MC chips from Q4 2027 to Q4 2026 to reflect stronger production ramp-up.

MediaTek faces more noticeable shipment adjustments. SemiAnalysis cumulatively lowered the lifetime shipments of Humufish by 470,000 units for 2028–2029 and reduced the lifetime shipment forecast for the Triggerfish project targeting Anthropic by 42%.

**Amazon's Trainium is one of the few platforms whose near-term forecasts were not downgraded in this adjustment.** SemiAnalysis did not cut the shipment forecasts for Trainium in 2027 and 2028, as Amazon has a sufficient reserve of data center projects, Trainium has lower manufacturing costs, and Amazon is willing to support its key customers through strategic subsidies.

Overall, SemiAnalysis's adjustment does not mean that AI compute demand will disappear entirely, but rather that growth expectations previously built on continuous financing, rapid data center expansion, and high-priced hardware are being recalibrated.

As AI lab revenue growth slows and capital market financing tightens, competition among chip suppliers will shift from pure performance comparisons to a comprehensive contest involving performance-to-cost ratios, energy efficiency, and financing capabilities.

NVIDIA still holds a strong position in the AI chip market, but in SemiAnalysis's view, its future growth will face greater pressure than that of some ASIC suppliers.

### Related Stocks

- [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md)
- [ANV.US](https://longbridge.com/en/quote/ANV.US.md)
- [DIPS.US](https://longbridge.com/en/quote/DIPS.US.md)
- [NVC.US](https://longbridge.com/en/quote/NVC.US.md)
- [NVD.US](https://longbridge.com/en/quote/NVD.US.md)
- [NVDB.US](https://longbridge.com/en/quote/NVDB.US.md)
- [NVDD.US](https://longbridge.com/en/quote/NVDD.US.md)
- [NVDG.US](https://longbridge.com/en/quote/NVDG.US.md)
- [NVDL.US](https://longbridge.com/en/quote/NVDL.US.md)
- [NVDO.US](https://longbridge.com/en/quote/NVDO.US.md)
- [NVDQ.US](https://longbridge.com/en/quote/NVDQ.US.md)
- [NVDS.US](https://longbridge.com/en/quote/NVDS.US.md)
- [NVDU.US](https://longbridge.com/en/quote/NVDU.US.md)
- [NVDW.US](https://longbridge.com/en/quote/NVDW.US.md)
- [NVDX.US](https://longbridge.com/en/quote/NVDX.US.md)
- [NVDY.US](https://longbridge.com/en/quote/NVDY.US.md)
- [NVIB.US](https://longbridge.com/en/quote/NVIB.US.md)
- [NVII.US](https://longbridge.com/en/quote/NVII.US.md)
- [NVIT.US](https://longbridge.com/en/quote/NVIT.US.md)
- [NVW.US](https://longbridge.com/en/quote/NVW.US.md)
- [NVYY.US](https://longbridge.com/en/quote/NVYY.US.md)
- [NYYY.US](https://longbridge.com/en/quote/NYYY.US.md)
- [07388.HK](https://longbridge.com/en/quote/07388.HK.md)
- [07788.HK](https://longbridge.com/en/quote/07788.HK.md)
- [09388.HK](https://longbridge.com/en/quote/09388.HK.md)

## Related News & Research

- [BUZZ-Street View: Nvidia's AI story intact despite uncertainty at top](https://longbridge.com/en/news/305945171.md)
- [Trump’s Nvidia stake hit $83 million, then disclosure gaps took over](https://longbridge.com/en/news/305781969.md)
- [Jim Cramer says Nvidia’s leadership shake-up won’t derail the AI trade](https://longbridge.com/en/news/305736094.md)
- [QUOTES-People react to US AI chip curbs that threaten data centre boom](https://longbridge.com/en/news/305767058.md)
- [Nvidia Reports $115.8B in Quarterly Revenue Under Interim CEO](https://longbridge.com/en/news/306103086.md)

---
> **Disclaimer: This article is for reference only and does not constitute any investment advice.**