UBS Sees Buying Opportunity After Tech Stock Plunge: Market Underestimates AI Supercycle, Inflection Point for Returns May Come in 2027

According to Zhihu Finance APP, after the Nasdaq 100 index—dubbed the "global AI trading and tech stock benchmark"—lost more than $2 trillion in a single month in October, Gerry Fowler, Chief European Equity Strategist and Head of Derivatives at UBS Group, said in an interview on Monday that after the recent retreat in AI trading, the valuations of hyperscale cloud service providers now appear attractive, as the market's latest pessimistic pricing for the AI super cycle reflects a duration shorter than what may actually materialize.

“When the overall valuation multiples of hyperscale cloud service providers such as Alphabet, Microsoft, and Amazon have fallen below 20 times, while their revenues are still growing by 10% to 20% annually, it indicates that the market has taken a rather pessimistic view. Therefore, I believe the willingness to buy these stocks on dips is quite strong,” he said in a media interview on Monday.

Fowler said hyperscale cloud service providers may be approaching the point where their massive AI expenditures begin to yield returns. He predicts that a turning point could emerge in 2027, when data center deliveries will drive the fulfillment of order backlogs, with revenues finally catching up to capital expenditures. He added that capital has not disappeared from the market, but has rather shifted to the core hardware providers at the heart of AI compute infrastructure—such as storage chip manufacturers—who are now beginning to return capital to investors.

The consensus expectation for trillion-dollar capital expenditures (CapEx) from the large tech giants or hyperscale cloud superpowers by 2030 is mainly being used to purchase AI chips, servers, and other resources needed to build large AI data centers and conventional cloud computing; AI financing costs primarily include borrowing interest, financing fees, etc., and are usually accounted for separately from CapEx, though a small part of eligible construction period interest may be capitalized and cannot be double-counted as total costs.

Alphabet, Google's parent company, reported last week that its free cash flow was negative for the second consecutive quarter, heightening pessimism and anxiety around the global AI trading theme. The Nasdaq 100 Index, which is heavily weighted toward tech stocks associated with AI themes, fell 3.5% on Friday, marking its largest single-day decline since June.

However, in Fowler's view, AI trading is beginning to expand beyond the first wave of data center beneficiaries. Many suppliers focusing on cables, electrical equipment, and other infrastructure for the construction of large AI data centers have already sold out their capacities for 2027, even 2028, narrowing the scope for upward earnings revisions. Because the market is driven by unexpected changes in growth, he believes the next phase of this unprecedented AI bull market must seek growth drivers from other key areas.

As shown above, Alphabet, Google's parent company, has burned cash for two consecutive quarters—AI compute infrastructure-related expenditures have pushed free cash flow, which had remained positive for twenty consecutive years, into negative territory.

He went on to point out that critical industrial firms worldwide and in Europe, which help build new factories, deserve attention. He particularly emphasized in the interview that lithography machine giant ASML (ASML.US) is a prime example: this company supplies key semiconductor equipment—state-of-the-art EUV and even high-NA EUV lithography machines—to memory chip manufacturers like Samsung Electronics and SK Hynix who are expanding production, rather than directly selling AI infrastructure hardware products to AI data centers.

“As a result, ASML has suddenly started to receive earnings forecast upgrades, and now a considerable number of key industrial firms in Europe and beyond are experiencing the same,” he said.

Global Tech Stocks Hit Hard: Market Is Repricing for the Wait for Cash Returns

UBS's Fowler's core bullish view on the global AI compute theme and on tech stocks in general is that the recent sell-off may have underestimated the duration of the AI investment cycle. In the perspective of this European chief strategist, the 2027 wave of AI data center deliveries is expected to drive backlogged orders into strong revenues, while the large-scale capacity expansion at storage chip makers and chip manufacturing powerhouses like TSMC may in turn significantly boost earnings growth for industrial system suppliers such as ASML, opening up new investment opportunities around the AI theme.

On October 29, the Nasdaq 100 index and the Philadelphia Semiconductor Index fell about 1.38% and 1.42% respectively; these indices then plummeted a further 3.46% and 7.20% on October 30. The Nasdaq 100 delivered its largest single-day drop since June, while Intel plunged 28.58%—its worst one-day performance since 1974.

The credit market sheds light on the source of pressure behind this sell-off. On October 15, CoreWeave’s five-year Credit Default Swap (CDS) spread topped 1,100 basis points, and the yield on bonds maturing in 2031 reached about 16%; the cost for default protection for Alphabet, Amazon, and Meta also hit new highs, but rose much less sharply than CoreWeave’s. Capital is raising its requirements for the payback period of AI investments and debt-servicing ability.

Alphabet’s free cash flow was minus $9.065 billion for Q3, negative for two consecutive quarters; Amazon saw a net free cash flow outflow of about $42.4 billion over the twelve months ending September. Over the same period, Google Cloud revenue grew 103% and AWS grew 39.3%, illustrating that the debate around AI trading is focused on whether strong demand can be quickly converted to cash returns covering investments. It’s important to note, however, that increased negative free cash flow boosts the need for external financing, but does not directly imply operating losses or the same level of credit deterioration as new cloud companies.

From Data Center Deliveries to Memory Expansion: AI Rally Looking for the Next Wave of Profit Growth

The turning point Fowler suggests for 2027 follows a clear industry logic: only after data centers complete power supply, server deployment, and acceptance can signed orders be converted into billable services; higher utilization is needed to improve capital returns.

Another key channel for growth runs from strong storage demand driving record wafer fab capacity expansions, in turn generating more orders for lithography, deposition, and etching equipment. The investment focus is to identify forecast trends that have not yet fully priced in optimistic and sustained earnings expansion, rather than simply assuming share prices will match order backlogs in appreciation.

From the perspective of colossal AI inference workloads, the ongoing back-end execution of globally popular AI agents like Muse, along with the improved capabilities of cutting-edge AI models like Astra in complex computing tasks, has extended the range of AI-applicable workloads; multi-step tasks, tool invocation, and parallel agent processes all increase per-user model calls and context management. Anthropic has reported that token usage for multi-agent tasks is about 15 times that of regular chat in its research systems. Based on this, the comprehensive penetration of advanced AI agents like Muse will drive AI compute demand along GPU processing, HBM and DRAM bandwidth and capacity, KV cache and SSD storage, high-performance CPU tool execution, and high-speed data transmission systems based on optical interconnect chips. Total resource requirements will ultimately depend on the interplay between task volume growth and efficiency improvements per task.

From an engineering standpoint, agents expand a single QA into planning, retrieval, tool invocation, code execution, and result verification, with multiple rounds of inference raising computational needs. Management at US semiconductor equipment giant Applied Materials has previously noted that agent applications urgently require massive AI GPUs and other AI accelerators, as well as more CPU-intensive computing architectures, and demand will grow long-term for DRAM and NAND, fueling extra momentum for wafer manufacturing equipment. This helps explain why “the world’s biggest chip foundry” TSMC posted Q3 revenue of $46.72 billion, up 41.2% year-on-year, with a gross margin of 66.4%, and forecasts total chip equipment-related investment from 2026–2028 to exceed the previous three years significantly.

Thus, the semiconductor equipment sector benefits from both “capacity expansion needs” and “process upgrades.” Advanced logic chips’ gate-all-around structures, higher-layer 3D NAND, HBM stacking, and advanced packaging all raise the technical demands on EUV and High-NA EUV lithography, precise deposition, selective etching, chemical mechanical polishing, and defect management. Persistent growth in demand may spur both new production lines and upgrades to existing facilities.

Morgan Stanley’s Wilson complements this analysis from a market-timing perspective: midterm elections, AI financing pressures, and liquidity constraints may first push the S&P 500 down to 7,100, but after risk absorption, he still looks to 8,000. He believes companies’ ability to shift between equity and debt financing helps them adapt to changing capital costs; high-quality firms are better able to withstand these conditions. Kobeissi’s statistics show AI stocks dropped over 13% in October, while the “AI-excluded” portfolio rose 8%, indicating sector rotation. USB’s focus is on the AI investment cycle being longer than initially expected, while Wilson highlights adjusted reentry opportunities; both are concentrated on whether earnings can withstand current financing pressures.