Wall Street has largely accepted the fact that Big Tech will continue pouring billions into artificial intelligence (AI). The bigger question now is how soon these massive investments will translate into meaningful returns. Investors are increasingly estimating the timeline for these investments to start paying off.
Hyperscalers Funding the AI Buildout: Chipmakers Benefit Before Big Tech
Leading cloud and technology companies — including Amazon, Alphabet, Meta Platforms, Microsoft and Oracle — are investing aggressively in AI infrastructure, from advanced chips and data centers to power capacity.
While these hyperscalers are bearing the upfront costs, semiconductor companies are benefiting first by supplying the essential hardware powering the AI boom, as pointed out in a Yahoo Finance article.
BofA Sees a 'Generational' Shift in Cash Flow
According to Bank of America Global Research, the AI investment cycle is driving a "generational transfer in free cash flow" for semiconductor companies, per the same Yahoo Finance article.
In fact, free cash flow — the cash left after operating expenses and capital investments — is moving in opposite directions for hyperscalers and chipmakers. Chipmakers receive immediate revenues by selling AI hardware, while hyperscalers must absorb the cost of building infrastructure before realizing returns.
BofA's analysis shows that free cash flow for major hyperscalers has slipped into negative territory as capital spending surges, whereas semiconductor companies such as NVIDIA, Micron, Broadcom and Applied Materials continue to generate increasing free cash flow.
The firm estimates that the "Magnificent Seven" hyperscalers have spent roughly $234 billion on capital expenditures this year, even as their stocks have remained largely unchanged in 2026. Roundhill Magnificent Seven ETF MAGS has added about 3.7% so far this year (as of July 10, 2026). On the other hand, VanEck Semiconductor ETF SMH has surged about 63.7% in the year-to-date frame (as of July 10, 2026).
Apollo Flags Timing Risk
Apollo chief economist Torsten Sløk argues that the biggest uncertainty is not whether AI will generate value, but how long it will take. He points to two major challenges.
First, declining token prices allow AI usage to expand without generating proportional revenue growth. Second, rapidly improving Chinese AI models are increasing competitive pressure on U.S. technology platforms seeking to monetize AI services, the same Yahoo source highlighted.
Chinese AI Models Gain Market Share
Sløk highlighted a growing shift in AI usage toward Chinese models.
Chinese AI models are rapidly capturingglobal marketshare, driven by a dramatic cost advantage and rapidly closing capability gaps. The share of tokens used by U.S. companies on Chinese AI models via OpenRouter — a platform that enables developers to access a range of AI models — has stayed above 30% each week since Feb. 8, with that figure touching as high as 46%, as quoted on CNBC.
In June, AI startup Lindy shifted 100% of its traffic from Anthropic’s Claude models to DeepSeek, a Chinese company that shook the market in early 2025 with the release of its low-cost AI models.
Boom-and-Bust Risks Remain for Chip Companies
Most chipmakers are signing multi-year supply agreements, requiring fixed purchase commitments and substantial upfront payments. However, while long-term contracts may provide greater revenue visibility, they may not totally eliminate the chip and memory markets’ cyclical nature.
Between 2022 and 2023, following a post-pandemic slump in PC and smartphone sales, memory manufacturers were selling products below production costs. Hence, despite the ongoing AI-led demand, it remains unclear whether supply discipline can prevent another boom-and-bust cycle once memory shortages ease.
Bottom Line
Despite near-term cash flow pressure, the long-term outlook for AI remains constructive if adoption continues to accelerate and businesses successfully monetize AI services. The Zacks Rank #1 (Strong Buy)State Street Technology Select Sector SPDR ETF XLK is still a good bet. XLK is up 0.5% over the past month (as of July 10, 2026).
Investors can also play the Roundhill Memory ETF DRAM. The DRAM is down currently as the memory market is adjusting to overvaluation. However, the fundamentals are still strong. DRAM is off 3% over the past month.
Boost Your Portfolio with Our Top ETF Insights
Zacks' exclusive Fund Newsletter delivers actionable information, top news and analysis, as well as top-performing ETFs, straight to your inbox every week.
Don’t miss out on this valuable resource. It’s free!
Get it now >>State Street Technology Select Sector SPDR ETF (XLK): ETF Research Reports
VanEck Semiconductor ETF (SMH): ETF Research Reports
This article originally published on Zacks Investment Research (zacks.com).
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
