Introduction: From Nvidia and Broadcom to optical modules and semiconductor equipment, AI hardware assets are undergoing an expected revaluation.

In the past two years, one of the most crowded transactions in the global capital market has been the purchase of AI hardware.
Nvidia’s GPUs, Broadcom’s custom chips, TSMC’s advanced processes, as well as optical modules, servers, power supplies and cooling equipment almost constitute the most certain beneficiary chain in the AI investment wave.
Investors’ logic at the time was not complicated: As long as Microsoft, Google, Amazon and Meta continue to increase capital expenditures, the demand for AI computing power will continue to grow, and hardware companies in the upper reaches of the industry chain will naturally be able to obtain orders, revenue and profits.
But by November 2026, the market’s trading logic is changing.
In the past, investors worried about not buying enough AI chips; now, they are worried about buying too many chips. In the past, the market was chasing order growth and production capacity expansion; now, more and more funds are beginning to examine capital return, customer financing capabilities, equipment utilization, and ultimately cash flow.
This does not mean that the long-term development logic of the AI industry has ended, but it means that global investors are re-pricing the AI industry chain.
In this change, the first to be affected are the AI hardware companies that have experienced the largest increase, the highest valuation, and the most consistent market expectations.
01 From buying AI hardware to scrambling to reduce exposure
This round of adjustments to AI hardware assets did not start with a sudden deterioration in a certain company’s performance, but rather with multiple market signals changing at the same time.
On November 5, U.S. semiconductor stocks suffered a violent sell-off.
The Philadelphia Semiconductor Index fell 12.2% in a single day, and the Nasdaq Index fell 5.5%. This is the largest one-day drop for the Philadelphia Semiconductor Index since April 2025.
Qualcomm’s financial report fell short of expectations and became a direct trigger for market adjustment, but what really makes investors uneasy is whether AI infrastructure investment can continue to expand at the previous rate.
Reuters quoted Mark Hackett, Nationwide’s chief market strategist, as saying that hedge funds had been actively selling stocks of semiconductor and new cloud computing companies for several consecutive weeks, and some funds still tended to go short or reduce leverage.
This shows that the market adjustment is not just short-term profit taking, but that some institutions have begun to actively reduce the risk exposure of the AI industry chain.
This change quickly spread to the Asian market.
On November 16, China’s A-share AI hardware sector once again experienced significant adjustments. The 5G Communications Index fell 7.5%, the Science and Technology Innovation 50 Index fell 5.2%, and Zhongji InnoLight fell 12.3% in a single day.
At the same time, semiconductor equipment companies in the Japanese market also suffered a sell-off. Tokyo Electron fell 7.1% and Advantest fell 11.1%.
From U.S. chip stocks, to Japanese semiconductor equipment, to Chinese optical module companies, the AI hardware industry chain has shown obvious cross-market resonance.
What is even more noteworthy is that institutional investors are beginning to re-examine how many AI assets they hold.
On November 16, Reuters reported that among the world’s major companies, the investment assets of companies directly affected by AI reached at least US$25.3 trillion. For many large investment institutions, AI is no longer an industry theme that can be managed alone, but a common source of risk deeply embedded in stocks, bonds and other assets.
The U.S. PGIM has even begun to conduct AI risk screening on the nearly US$1 trillion of fixed income assets it manages, not only to assess the extent to which corporate business models are affected by AI, but also to assess the impact that the slowdown in AI capital expenditures may have on corporate credit status.
The logic behind this is that if there is a correction in AI investment, it may not only be affected by chip companies, but also data center developers, equipment suppliers, power companies, and even financial institutions that provide financing for these projects.
Of course, it is not accurate to interpret this change as global investors have unanimously been bearish on AI hardware.
For example, an analysis of global client transaction data by Saxo Bank, a subsidiary of Julius Baer, shows that after semiconductor stocks plummeted on November 5, some clients actively bought stocks in companies such as Nvidia, AMD, Micron and SanDisk.
Among them, NVIDIA’s daily average buying transaction activity has increased significantly compared with before.
This shows that the market has not given up on AI hardware, but is re-selecting the price and objects to buy.
02 The most dangerous changes in AI hardware
If the market adjustment in early November reflects changes in investor sentiment, then the latest AI accelerator forecast model released by SemiAnalysis on November 13 provides an industry-level basis for this adjustment that deserves more attention.
The research institution, which has long tracked global semiconductor and AI infrastructure, has significantly lowered its forecasts for global AI accelerator capital expenditures, chip shipments, HBM demand and power demand in the next few years.
Among them, the most eye-catching is the global AI accelerator capital expenditure forecast.
According to its latest model, the global AI accelerator capital expenditure forecast in 2027 is lowered from the previous US$770 billion to US$627 billion, a decrease of 19%; in 2028 it is lowered from US$1.054 trillion to US$758 billion, a decrease of 28%; in 2029 and 2030 it is lowered by 41% and 40% respectively.
By 2030, global AI accelerator capital expenditure forecast will drop to US$923 billion from the previous US$1.531 trillion, a decrease of US$608 billion.
This is not a small forecast revision.
It means that the past market expectations for continued rapid growth in AI computing power investment are being recalculated.
Of course, this doesn’t mean AI infrastructure investment will stop growing. According to SemiAnalysis’s adjusted model, global AI accelerator capital expenditure will still grow from approximately US$392 billion in 2026 to US$923 billion in 2030.
The question is, at what growth rate have investors previously valued these companies?
When the scale of future investment still grows, but the growth rate is significantly lower than originally expected, the valuation of hardware companies needs to find a new basis.
This is especially true for companies whose share prices already reflect rapid growth in the coming years.
SemiAnalysis’s forecast adjustment also involves a more important change: the price and profit margin of hardware products.
In the past two years, the supply of AI chips has been tight, and HBM memory, high-end packaging and advanced process production capacity have been in relative shortage.
In an environment where supply exceeds demand, suppliers can not only obtain more orders, but also obtain higher profits by increasing product selling prices and optimizing product structure.
But as more capacity enters the market, supply and demand are changing.
SemiAnalysis lowered its global HBM shipment forecast in 2028 from the previous 9.0EB to 6.4EB, a decrease of 30%. It also believes that HBM supply and demand are improving and future price assumptions need to be lowered.
For chip suppliers, this means a new problem: even if shipments continue to grow, average selling prices and profit margins may not be able to maintain previous levels.
The impact on Nvidia is particularly noteworthy.
SemiAnalysis lowered Nvidia’s computing revenue-related shipment forecasts by 22%, 39%, 56%, and 66% respectively from 2027 to 2030, and lowered the average selling price assumptions of some new-generation products.
Behind this is not only the slowdown in demand growth, but also changes in the competitive landscape.
In the past, NVIDIA GPUs were almost the first choice for large-scale AI model training and inference. But now, custom ASICs developed by Google TPU, Amazon Trainium and Broadcom are becoming another option for large AI companies.
For giants like Microsoft, Google, Amazon and Meta, they don’t necessarily need to rely on the same chip architecture forever.
As long as other options can provide better performance-cost ratio, lower energy consumption or more controllable supply chain, they have the incentive to adjust the procurement structure.
What this means for Nvidia is that even if global demand for computing power continues to grow, it may not be able to achieve the same market share and profit margins as in the past.
In contrast, Broadcom has performed relatively steadily in this forecast adjustment. SemiAnalysis lowered its ASIC business cumulative revenue forecast by 11% from 2026 to 2030, mainly related to changes in HBM cost and product price assumptions, rather than large-scale shipment reductions.
This also explains why the AI hardware industry chain cannot simply be viewed as a whole.
Different companies’ competitive advantages, customer structures and product positioning are determining their performance in the next cycle.
However, what really worries the market is not just competition among chip suppliers.
The deeper question is, who will continue to pay for this AI infrastructure construction?
In the past, the market believed that large cloud computing companies had strong enough cash flow to support AI capital expenditures for many years.
But now, the scale of investment has become larger and larger.
The global investment strategy report released by JP Morgan Private Bank in November pointed out that in the third quarter of 2026, the capital expenditures of three large cloud computing companies were lower than market expectations, and the upward revision of capital expenditure forecasts for the next 12 months also fell to the lowest level since the fourth quarter of 2023.
This does not mean that Microsoft, Google, Amazon and Meta will stop investing in AI, but it does mean that these companies will start to pay more attention to the pace of investment and financial discipline.
A study released by the Federal Reserve Bank of St. Louis in November also revealed the changes in the capital structure behind this round of AI investment.
The study found that the cash-to-asset ratio of large cloud computing companies fell from 35% in 2018 to 16% in 2026, while at the same time, their invested-to-asset ratio nearly doubled.
What is even more noteworthy is that there is a clear difference in the financing status of data center construction companies and semiconductor companies.
From the fourth quarter of 2024 to the second quarter of 2026, the quarterly average net debt financing scale of data center construction companies increased from US$3.23 billion to US$30.17 billion; while semiconductor companies shifted from net debt to net debt repayment.
This means that the financing pressure of the AI industry chain is not evenly distributed.
Chip suppliers are still able to rely on strong operating cash flow to support business development, but some data center builders and new cloud computing companies need to increasingly rely on external financing.
Once the financing environment tightens or customers are no longer willing to commit to purchases in advance, some data center projects may be delayed, thereby affecting orders for chips, servers, network equipment and power equipment.
This is why the market has begun to re-examine the order quality of the AI industry chain.
In the past, as long as the customer announced the construction of a data center and signed a purchase agreement, the supplier could gain market recognition.
Now, investors are more concerned about: Do customers have enough funds? When will the project be delivered? Can the equipment be put into use? Will the final revenue cover depreciation, interest and operating costs?
These issues are replacing simple order growth as an important factor in determining the valuation of AI hardware companies.
03 The era of buying with eyes closed is over
The real difficulty for global AI hardware companies may not be the immediate decline in stock prices, but changes in the way the market evaluates them in the coming quarters.
In the past, the most important competitive advantage for AI hardware companies was their ability to enter the supply chains of Nvidia, Microsoft, Google, Amazon and Meta.
As long as it enters the supply chain and obtains orders, the market is willing to give a higher valuation.
But now, supply chain identity is no longer a sufficient condition.
Among AI chip companies, some have long-term purchasing commitments from large cloud computing customers, while others rely on new cloud service providers with weak financing capabilities; among optical module companies, some products are in a technology upgrade cycle, while others may face pressure from customer concentration, price competition, and slowdown in capital expenditures.
These differences will gradually be reflected in the company’s revenue growth, gross profit margin and cash flow.
Chinese AI hardware companies face the same problem.
In the past year, AI hardware transactions in the A-share and Hong Kong stock markets have largely benefited from the growth in capital expenditures of overseas cloud computing giants.
Although companies such as Zhongji InnoLight, Xinyi Sheng, and Tianfu Communications are in different parts of the industrial chain, their market pricing is closely related to the construction of global AI data centers.
If overseas cloud factories continue to expand capital expenditures, Chinese suppliers will have the opportunity to obtain more orders; but if the pace of global data center construction slows down, or customers begin to adjust equipment purchase plans, it will be difficult for China’s supply chain to completely stay away.
The sharp drop in Zhongji InnoLight’s stock price on November 16 is a microcosm of the risk transmission of this global industrial chain.
Of course, China’s AI industry also has its own development logic.
The construction of domestic AI chips, domestic servers and data centers is forming an independent demand market. Restrictions on overseas chip supply may also promote the accelerated localization of domestic computing infrastructure.
However, domestic substitution does not mean that changes in the global AI investment cycle can be completely hedged.
For hardware suppliers, they still have to answer the same question in the end: do customers have real needs, does the project have continued financial support, and can the product generate stable profits.
From this perspective, what global investors are doing is not a simple retreat from the AI industry, but a switch in investment logic.
In the first stage, what the market buys is the scarcity of computing power.
In the second stage, what the market buys is order growth.
Next, the market may be more willing to pay a premium for companies that can create real cash flow, have stable customers and strong pricing power.
This also means that the differentiation within the AI hardware industry chain may further increase.
Companies with technical barriers, customer stickiness and cost advantages may still benefit from the next round of industrial expansion; but those companies that mainly rely on industry prosperity, customer financing and high capital expenditure expectations to support their valuations will need to face more stringent market testing.
For investors, what needs to be paid attention to in the future is no longer just how much NVIDIA’s next-generation GPU can be sold, or how much global demand for optical modules can grow.
What is more important is to observe whether the capital expenditures of large cloud computing companies continue to increase, whether AI application revenue can cover computing power costs, whether data center utilization can be improved, and whether hardware suppliers can maintain profit margins in price competition.
If these indicators cannot continue to improve, the stock prices of some hardware companies may continue to be under pressure even if the AI industry is still in a growth cycle.
The AI revolution will not end with a market correction.
However, the way the capital market prices AI hardware has changed.
In the past, investors believed that as long as AI continued to develop, shovel sellers would make money.
Now, they began to ask another question: If everyone is buying shovels, how many people will eventually be able to make money from mining?
Related Reading
- Anthropic engineer’s warning about Claude Code explains why tech outages could get much worse2026-11-22
- Original Unexpectedly, just one day after the APEC summit, China became famous around the world because of its three “specialties”2026-11-21
- Original Anthropic goes public and creates wealth: 96% of employees became millionaires overnight2026-11-20
- The beauties and handsome guys from UB.com want to enter your home and wait until they arrive.2026-11-20
- Original: Tencent’s shareholders will suffer again2026-11-20