· Based on the recent AI developments and related companies’ actions and plans, we still believe that the long-term structural story of AI remains intact.
· Earnings growth of semiconductor companies with significant exposure to AI infrastructure are expected to remain strong, supported by sustained AI CAPEX.
· While there are signs of a valuation bubble forming in semiconductor stocks, the current market is different from the dot-com bubble: (1) AI CAPEX are still being largely funded from free cash flow, as opposed to debt. Currently, greater risk lies in under-investment in AI rather than long-term overbuilding; (2) valuations are still well below past extremes.
· On the other hand, the current pace of growth in CAPEX spending appears unsustainable, with markets currently expecting more aggressive investments ahead. Free cash flows have so far supported the boom, but maintaining this aggressive CAPEX could eventually strain balance sheets.
· Semiconductor stocks are currently priced to perfection, leaving little room for error. A continued rally will largely depend on earnings exceeding the already-high expectations, while any disappointments could trigger a sharp pull-back in share prices. Yet, even if there is a valuation bubble, it is difficult to tell when the bubble will burst
· A downgrade of the semiconductor sector outlook to 2.5 stars “Neutral” is warranted. Investors who bought earlier may consider taking profit, while those still on the sidelines should avoid taking large new positions; Long-term bulls on the sector may consider building exposure to large-cap AI-enablers or through a regular savings plan (RSP).
Since our article in early September (Downgrade Semiconductor to 3.0 Stars: Focus on AI Enablers), the MVIS US Listed Semiconductor 25 Index has risen over 26% (as of 30 Oct 2025) in less than two months, exceeding our earlier target price.
This rapid achievement is not only due to market expectations that hyperscalers, AI startups, and sovereign nations will maintain strong CAPEX on AI infrastructure in the foreseeable future, but more so attributable to OpenAI's recent efforts to promote an "AI circular economy" in collaboration with hyperscalers and semiconductor giants, optimistic market expectations for tech giants' earnings in 3Q25, and positive revenue signals released by Nvidia’s CEO Jensen Huang at the latest GTC conference—all of which have fuelled the market's continued enthusiasm for AI.
However, amid the stock market frenzy, many investors remain skeptical of the high valuations of large tech companies and semiconductor manufacturers, the sustained large-scale AI investments, and the formation of the AI closed-loop supply chain, once again comparing the current AI era to the dot-com bubble of 2000, reigniting talk of an "AI bubble."
As a matter of fact, based on the recent AI developments and related companies’ actions and plans, we still believe that the long-term structural story of AI remains intact. However, we are convinced that early signs of an AI bubble are emerging, and investors should adopt a more cautious approach to the semiconductor sector to better manage potential risks. This article will explain the underlying reasons.
Comparing AI and the 2000 Dot-Com Bubble
Figure 1: Cisco vs. Nvidia Valuation
The chart above compares the valuations of Cisco during the dot-com bubble and Nvidia today. As the highest-market-cap tech companies in their respective eras, Cisco's forward PE multiple reached over 100x at its peak in March 2000 (the bubble's peak), while Nvidia's current forward PE multiple is around 30x. This indicates that Cisco's pricing at that time reflected far more speculative earnings growth than what we see today, whereas Nvidia's market cap, despite multiplying in recent years, is clearly supported by robust earnings growth, highlighting a significant difference in their profitability.
Figure 2: Magnificent 7 vs. Tech Bubble Leaders' Valuations and Fundamentals
| Tech Bubble Leaders | Forward P/E (2000) | Return on Equity (ROE) | Net Income Margin (NIM) |
| Microsoft | 53.2 | 35.0% | 39.0% |
| Cisco Systems | 101.7 | 22.0% | 17.0% |
| Intel | 42.1 | 26.0% | 25.0% |
| Oracle | 84.6 | 39.0% | 15.0% |
| IBM | 23.5 | 39.0% | 9.0% |
| Lucent | 37.9 | 36.0% | 9.0% |
| Nortel Networks | 86.4 | -1.0% | -1.0% |
| Average | 61.3 | 28.0% | 16.1% |
| Magnificent 7 | Forward P/E (2026) | Return on Equity (ROE) | Net Income Margin (NIM) |
| Nvidia | 27.0 | 119.2% | 55.8% |
| Meta | 17.6 | 37.1% | 37.9% |
| Microsoft | 26.6 | 33.3% | 36.2% |
| Amazon | 20.1 | 24.3% | 9.3% |
| Google (Alphabet) | 21.5 | 32.9% | 28.6% |
| Apple | 30.0 | 171.4% | 26.9% |
| Tesla | 145.0 | 10.5% | 7.3% |
| Average (excl. Tesla) | 23.8 | 69.7% | 32.5% |
Comparing the current Magnificent 7 with seven tech giants from the dot-com bubble era, the former (excluding Tesla) have average forward PE multiples lower than the latter, but their fundamentals (measured by ROE and NIM) are clearly superior. This suggests that compared to companies during the dot-com bubble, today's dominant tech companies have lower valuations but stronger earnings growth. The same results apply when comparing the top seven components of the MVIS US Listed Semiconductor 25 Index (including Nvidia, TSMC, Broadcom and ASML) with the seven dot-com bubble tech giants.
However, within the broader AI sector, excluding large companies with actual profitability, there are clear deviations in valuations and fundamentals for certain tech firms, such as Palantir and Nebius. Although the former is projected to have double-digit earnings growth in the coming years, its forward PE multiple exceeds 100x, while the latter's stock price has surged despite remaining unprofitable. Both have seen their share prices multiply several times over the past year, suggesting that market pricing for these AI concept stocks may reflect irrational over-optimism detached from fundamentals. Overall, we believe we are currently in the early stages of an AI bubble.
Sustainability of Tech Giants' AI CAPEX in Doubt
While the fundamentals of hyperscalers and large semiconductor manufacturers remain strong currently, the projected CAPEX by hyperscalers on AI infrastructure, alongside the recent AI circular economy formed by both and large model developers such as OpenAI, are raising red flags for our view on AI-related industries.
Figure 3: Hyperscalers'
Consensus CAPEX
Source: Bloomberg
Figure 4: Hyperscalers'
Free Cash Flow to CAPEX Ratio
Source:
Bloomberg and iFAST Financial compilation. Data as of 30 Oct 2025
From Figure 3, analysts expect the six largest hyperscalers' combined CAPEX to reach over USD 390 billion this year, more than double that of 2023. Following earlier upward revisions by Alphabet and Meta to their respective CAPEX targets and lower bounds, recent earnings reports from Amazon and Meta announced higher-than-expected CAPEX and further raised lower bounds respectively, indicating unabated demand for data centers and electricity. In this regard, AI CAPEX are expected to reach new historical highs next year. Furthermore, referencing Jensen Huang's prediction that global AI data center spending will reach USD 3-4 trillion by 2030, while we believe this figure may be exaggerated, we anticipate that large-scale AI investment will continue in the future.
However, while most of hyperscalers' current CAPEX come from free cash flow, some companies have begun using debt issuance for financing (e.g. Oracle's USD 18 billion bond and Meta's USD 25 billion bond). This indicates leverage is emerging in the AI ecosystem for large tech companies, increasing the necessity for them to achieve revenue and earnings targets to justify their need for capital, reminiscent of the dot-com bubble.
In fact, besides debt financing, Figure 4 shows that hyperscalers' free cash flow to CAPEX ratio has been declining continuously over the past few quarters. Although they still have sufficient leverage space in the foreseeable future, the downward trend indicates a clear shift in how companies manage their liquidity: the decline in cash flows and growing reliance on debt financing reflect changes in large tech companies' capital management priorities, which warrant investor attention.
In addition to investments from tech giants, in the private market, according to the Financial Times, the combined valuation of ten unprofitable global AI startups (including OpenAI, Anthropic, Perplexity, and Scale AI) has surged by nearly USD 1 trillion over the past year (as of Oct 2025), marking the fastest growth in history. These ten companies have attracted over USD 200 billion in venture capital (VC), accounting for two-thirds of the total US VC investments for the year.
Meanwhile, according to EY's analysis of Crunchbase data, AI-related investments accounted for 71% of VC in 1Q25, significantly higher than the 45% and 26% of the previous two years. This indicates the unabated investor enthusiasm for both the AI sector and the aforementioned loss-making AI companies, while their future monetisation capabilities remain uncertain, posing risks to the investment logic of funding AI build-out.
Potential Risks Behind the AI Closed-Loop Economy
Figure 5: Increasingly
Cyclical Ecosystem in the AI Field
Source: Goldman Sachs
Regarding the recent heated market discussion on the "AI closed-loop economy", in short, our view is that in the short to medium term, the revenue growth for semiconductor companies driven by the AI cyclical ecosystem, alongside the AI boom fuelled by vendor financing, may persist. However, in the longer term, while we believe demand for AI and high-performance computing will remain strong, given the difficulty in estimating the actual demand for AI computing power, the lingering risk of long-term infrastructure overcapacity, the currently low visibility of the actual monetisation capabilities of large-scale model developers, as well as the potential for increased complexity from further extensions of the AI cyclical ecosystem, we adopt a cautious outlook.
In fact, within this AI cyclical ecosystem, OpenAI has established partnerships with several large semiconductor companies and hyperscalers, enabling semiconductor manufacturers to strategically secure new chip orders. Meanwhile, the increase in AI workloads heightens demand for hyperscalers' cloud services, while OpenAI gains priority access to chips and computing infrastructure, creating mutual benefits for all three parties. Hence, the "AI closed-loop economy" has recently further boosted market optimism for AI, leading to surges in stock prices for large-cap AI enablers and hyperscalers, as well as OpenAI's valuation.
Figure 6: Triangular
Relationship Between OpenAI, Nvidia, and Oracle
Source: Wallstreetcn.com
Nevertheless, behind the short- to medium-term advantages lie risks to the long-term sustainability and interdependence of the tripartite collaboration. For example, in the cycle where Nvidia invests up to USD 100 billion in OpenAI to build at least 10 gigawatts (GW) of AI data centers, and Oracle procures GPUs from Nvidia, while we believe this partnership can accelerate data center build-out and lock in capacity in the short to medium term, in the longer term, should AI adoption rates fall short of expectations, leading to lower-than-expected monetization and revenue, the ripple effects from this AI closed-loop supply chain could negatively impact Nvidia's semiconductor sales and investment returns. In fact, vendor financing (where Nvidia provides equity investments to clients, who then use those funds to repurchase equipment) was a characteristic of the 2000 dot-com bubble, further reinforcing our view that the AI bubble is in its early stages.
Regarding the future practical applications of AI, indeed, we remain convinced that AI's long-term global impact will be structural and disruptive. We believe humans can leverage AI to automate repetitive tasks for efficiency, analyze large datasets to identify patterns and predict trends, and optimize complex processes for intelligent decision-making. Comparing the number of tokens processed by several AI companies (Google, Microsoft, and OpenAI) as of 2Q25, all have grown at a faster pace YTD, reflecting user growth and increased demand for inference models, which have partly contributed to their AI revenues.
Having said that, taking OpenAI as an example, since the launch of its large language model (LLM) ChatGPT at the end of 2022, the company still projected approximately USD 5 billion in losses last year and expects hundreds of millions in losses this year, with market expecting that it may not turn profitable until 2029. For revenues from LLMs, we believe application programming interfaces (APIs) and other consumption-based contracts will become the main drivers of LLM licensing sales in the medium to long term, but in the short term, we still consider this goal challenging to achieve. In other words, although tech giants and large model developers are gradually realizing AI revenues, they are currently far from catching up to the massive AI expenditures.
In light of this, given the still relatively low AI revenues of large-scale AI developers, we take a more cautious outlook. In the long term, we tend to believe that only Artificial General Intelligence (AGI) can justify the trillion-scale AI CAPEX for mega-scale computing in the coming years, but the likelihood of achieving AGI remains low in the foreseeable future, so uncertainty surrounding the overall monetisation capabilities of AI is expected to persist.
Potential risks leading to increased volatility in the semiconductor sector
Corporate earnings releases and guidance: We believe semiconductor stocks are currently priced to perfection, leaving little room for error. A continued rally will largely depend on earnings or guidance exceeding the already-high expectations, while any disappointments could trigger a sharp pull-back in share prices.
Future depreciation expenses: Considering that hardware like GPUs used for AI computing faces faster technological iterations and heavier workloads, their useful life may be only three to five years. The market may currently be underestimating hyperscalers' future depreciation costs, and increased depreciation expenses could drag on earnings and AI CAPEX.
Geopolitical developments: US-China trade tensions, strict export controls, and escalating military activities in the Taiwan Strait could lead to temporary disruptions in the supply of specific chips and reshape the global semiconductor supply chain.
AI revenues are overly concentrated in tech giants: It is expected that the development of AI in the next five years will mainly revolve around the B2B sector, with AI factories (leasing services) potentially becoming a growth engine. However, any revenue or visibility below expectations could reduce AI CAPEX and drag down the earnings of semiconductor companies, whose earnings mainly come from a few tech companies with large CAPEX.
Downgrading the 2026 semiconductor sector outlook
In summary, while there are signs of a valuation bubble forming in semiconductor stocks, we believe the current market is different from the dot-com bubble: (1) AI CAPEX are still being largely funded from free cash flow, as opposed to debt. Currently, greater risk lies in under-investment in AI rather than long-term overbuilding; (2) valuations are still well below past extremes.
However, after two years of AI development, the environment now seems to be quite different—the semiconductor sector's valuations have risen to higher levels, some hyperscalers have begun issuing debt for financing, and the AI closed-loop economy is emerging.
Meanwhile, we believe the long-term structural story of AI remains intact, and we expect strong earnings growth for the large-cap AI enablers (large companies with significant exposure to AI infrastructure) in the semiconductor sector.
In view of this, after weighing recent AI developments and company fundamentals, given a more limited upside potential and the difficulty to tell when the bubble will burst, we have decided to downgrade the star rating of the semiconductor sector from 3.0 stars "Attractive" to 2.5 stars "Neutral".
For investors who have bought earlier, they may consider to take profit by reducing positions at current historical highs; for those still on the sidelines, they may consider avoiding taking large new positions; for long-term bulls on the semiconductor sector, they may consider gradually building exposure to large-cap AI enablers, or to the sector via the VanEck Semiconductor ETF (NASDAQ:SMH) through a regular savings plan (RSP).
Table 1: MVIS US Listed Semiconductor 25 Index’s Earnings and Potential Upside
| 2024 | 2025E | 2026E | 2027E | |
| Earnings Per Share (EPS) | 287.10 | 433.4 | 556.9 | 669.2 |
| Earnings Growth YoY | 20.3% | 50.9% | 28.5% | 20.2% |
| PE Ratio (X) | 51.5 | 34.1 | 26.6 | 22.1 |
| Upside Potential (based on fair PE Ratio of 24X) | - | - | - | 8.6% |
| Source: Bloomberg Finance L.P., iFAST Compilations. Data as of 31 Oct 2025 | ||||
Declaration:
For specific disclosure, at the time of publication of this report, the analyst who produced this report and IFPL (via its connected and associated entities) holds a NIL position in the abovementioned securities.
All materials and contents found in this site are strictly for general circulation and informational purposes only and should not be considered as an offer, or solicitation, to deal in any of the funds or products found/identified in this site. While iFAST Financial Pte Ltd ("IFPL") has tried to provide accurate and timely information, there may be inadvertent delays, omissions, technical or factual inaccuracies and typographical errors. Any opinion or estimate contained in this report is made on a general basis and neither IFPL nor any of its servants or agents have given any consideration to nor have they or any of them made any investigation of the investment objective, financial situation or particular need of any user or reader, any specific person or group of persons. You should consider carefully if the products you are going to purchase are suitable for your investment objective, investment experience, risk tolerance and other personal circumstances. If you are uncertain about the suitability of the investment product, please seek advice from a financial adviser, before making a decision to purchase the investment product. Past performance is not indicative of future performance. The value of the investment products and the income from them may fall as well as rise. Opinions expressed herein are subject to change without notice. In respect of any matters arising from, or in connection with the said research analyses or research reports, recipients of the report are to contact IFPL at 10 Collyer Quay, #26-01 Ocean Financial Centre Building, Singapore 049315, or by telephone at +65 6557 2853. Where the report contains research analyses or research reports from a foreign research house and if the recipient of such research analyses or research reports is not an accredited investor, expert investor, institutional investor or an ex-accredited investor, IFPL accepts legal responsibility for the contents of such analyses or reports to such persons only to the extent as required by law. Please note that only certain security(ies) herein are available to all investors, while the rest are only available for certain persons to invest in, such as Accredited Investors (as defined in the Securities and Futures Act) or one who invests at least S$200,000 (or its equivalent currency) per transaction. To qualify as an Accredited Investor, one needs to submit a declaration form and certain relevant supporting documents, according to iFAST’s prevailing policies and procedures.
Please read our full disclaimers on the website at ( https://secure.fundsupermart.com/fsmone/policies/328125/investment-account-terms-&-conditions).
iFAST Financial Pte Ltd (IFPL) (registered address: 10 Collyer Quay #26-01 Ocean Financial Centre Singapore 049315, Telephone: 6557 2000) holds the Financial Advisers Licence issued by the Monetary Authority of Singapore ('MAS') to conduct regulated activities of advising on securities, marketing of collective investment schemes and arranging of any contract of insurance in respect of life policies, other than a contract of reinsurance and the Capital Markets Services Licence issued by the MAS to conduct regulated activities of dealing in securities and providing custodial services for securities. While IFPL has made every effort to ensure the independence of the report's contents, IFPL's nature of business is such that IFPL and its connected and associated entities together with their respective directors, officers and staff may be involved in providing dealing or investment-related services in the abovementioned securities, and have taken or may take positions in the securities mentioned in this report, and may also act as the principal for any buy or sell trades.