Shaken, Not Broken

Markets struggled to find their footing again this week as a sharp selloff in AI equities overshadowed an otherwise resilient macro backdrop. Through Thursday, the S&P 500 is −2% while the Nasdaq 100 was −3%, with most of the weakness concentrated in semiconductors and AI infrastructure names. The primary catalyst was growing investor concern around hyperscaler capital expenditures, which intensified after Alphabet raised its 2026 spending +8% to $200bn (mid) while also reporting its first cash burn as a public company. Investors increasingly questioned whether the pace of AI infrastructure spending can continue without a corresponding acceleration in monetization. Rising geopolitical tensions in the Middle East also weighed on sentiment, pushing Brent crude above $100/bbl, lifting Treasury yields, and reviving inflation concerns. Year-to-date, the S&P 500/NDX are still +8%/+12%. Our NPM Price data, which tracks the average price performance of the 50 largest names in our internal Tape D® database is +62%. (Wall Street Journal, 7/23/26; Bloomberg; NPM)

Is the AI Trade entering a new phase?

We are private markets people, but occasionally the public markets speak and we must listen. In public markets, from late 2022 through 2025, investors largely rewarded companies for ambitious AI strategies and increasing infrastructure spending. However, cracks in that foundation have emerged. Since peaking on 5/29, our index of hyperscalers (AMZN, GOOG, META, MSFT, ORCL; see chart below) is −12% (vs. −1% for the S&P 500), and our index of AI chip companies (AMD, AVGO, MRVL, MU, NVDA) is −13% (vs. 0% for the S&P 500) from a 6/22 peak. The market focus has shifted from “who is spending on AI?” to “who is earning attractive returns on those investments?”. For example, on 7/22, GOOG announced earnings, in which they increased their 2026 capex +8% and the stock fell 6%. (Reuters, 7/23/26; Financial Times, 7/23/26; Alphabet 7/22/26)

In our view, recent weakness across many AI equities has not been driven by fundamental demand changes, but, rather, it has reflected changing investor expectations. Concerns have emerged that hyperscaler capex may begin to normalize after the current infrastructure buildout, which, combined with high valuations and crowded positioning among institutional investors, is a cocktail for volatility. UBS forecasts that after +76% y/y capex growth in 2026, growth will slow to +26% in 2027 and +6% in 2028. (NPM; Reuters, 7/17/26; UBS 7/17/26; Deepwater Asset Management, 7/21/26)

The current AI investment cycle differs from recent, software-driven, technology booms because it encompasses virtually every layer of the computing stack. The first phase rewarded semiconductor manufacturers, particularly GPU suppliers. The second phase broadened to include infrastructure more broadly: networking equipment, optical components, advanced memory, power infrastructure, liquid cooling, data center construction, and cloud providers. More recently, investors have shifted attention toward application companies capable of monetizing generative AI. In short, hundreds of billions, if not trillions, of dollars are being spent on AI infrastructure, and the capital intensity has now led to increasing investor scrutiny on returns. (NPM; Reuters 7/15/26)

Who are Recent Winners and Losers?

Semiconductor companies have generally offered the most “AI beta” because they represent the purest expression of AI infrastructure spending. AMD, NVIDIA, Broadcom and Marvell ran up the most during the rally and therefore experienced disproportionate selling recently as investors reassessed future capital spending assumptions. By contrast, hyperscalers have held up relatively well. Investors increasingly believe that cloud providers will capture a disproportionate share of AI economics because they own the customer relationships, recurring revenue streams, and distribution channels needed to monetize AI at scale. (Bloomberg; NPM; Reuters, 7/17/26; 7/22/26; FT 7/23/26)

Memory suppliers and networking companies have generally remained resilient because current supply/demand fundamentals continue to support pricing and order visibility. Demand for high-bandwidth memory, networking equipment and advanced optical components remains exceptionally strong. (NPM; Bloomberg; Reuters, 7/17/26; 7/22/26; FT 7/23/26)

The current correction appears more consistent with a normalization of expectations than the unwinding of the AI investment thesis. The secular drivers remain intact: enterprise adoption continues expanding, cloud providers continue investing aggressively, governments increasingly view AI leadership as strategically important, and demand for accelerated computing remains robust. (Reuters, July 23, 2026; Financial Times, July 23, 2026; Alphabet Q2 2026 Earnings Call)

What are the Implications for Private Markets?

The recent volatility across public AI equities has implications for private markets. Historically, similar periods of public-market repricing have been associated with a widening dispersion between companies and increased investment discipline, trends we are monitoring closely in the private AI market. After nearly three years in which investors rewarded almost any company with perceived AI exposure, the market is increasingly differentiating between winners with durable competitive advantages and losers whose valuations were driven primarily by enthusiasm.

While there has been significant pressure on public semiconductor, infrastructure, and software names, the implications for private companies are more nuanced. Since the public AI trade began showing jitters, around 6/1, our basket of public AI stocks (the semiconductor and hyperscaler names mentioned above) is −8% and the S&P 500 is −1%, but our NPM Private Market Tracker (the average of the top 50 names by EV) is +10%. Even stripping out extreme outliers like BaseTen Labs (+139%) and SambaNova Systems (+145%) the average is still +5%. Bellwethers OpenAI and Anthropic were +3% and +11%, respectively. The chart below shows performance by sector in private markets since 6/1.

We note that, historically, private markets lag public markets by several quarters, and private investors rely somewhat on public market comparables when pricing late stage financings. Thus, there is a clear feedback loop between the two markets. However, we would note that, with the exception of the large LLMs, most private technology companies are focused on innovation, not infrastructure spending. As a result, we would expect the private markets to be more insulated from the primary concern nagging the public markets, which appears to us to be capital intensity and uncertainty about future cash returns.

YTD Private Market Performance by Sector