AI & Software Market Review

The AI trade continues but there is a rotation among the sectors. The semi sector is up over 700% since the launch of ChatGPT in late 2022, leading all sectors (P. 6), but was down a modest 3% in Q3 even with rising estimates. Multiples have compressed over this period even with robust revenue and earnings growth as the traditionally cyclical sector experiences some concerns on sustainability, perhaps relating to growing public resistance to data center capacity expansion in the US. Software has been mixed with the traditional SaaS sector out of favor, particularly in applications, with pressure during the “SaaSpocalypse” from October 2025 to April 2026 as investors expressed apprehensions of exposure to disruption from the new generation of AI native technologies. The applications sector rebounded a sharp 34% in Q3, but is still down 15% YTD from declines in the first half. Strong performance in Q3 came from Hyperscalers up 16%, Data and Dev Infrastructure up 24%, and Cybersecurity up 27%. The AI infrastructure and data center enablers was down 8% along with the semi sector’s 3% decline.

Rogue Agents. News of agents pursuing their narrow objectives and disregarding model constraints (going rogue) was a benefit to the security sector and may have benefited some of the applications companies as it appeared that AI native approaches may be slowed. A layered approach to containment and guardrails is being developed by various vendors and a consortium of companies coordinating with Nvidia is designed to provide security within the server itself, but layers of controls outside including the gateways and access to the internet will contribute further controls so the ultimate adoption of agentics is likely to progress but under scrutiny.

Rates are rising, but AI is the part of the economy least likely to slow. The Brookings Institute estimates that the US investment in datacenters and AI infrastructure will amount to $10.3 trillion over the 2025- 2032 period, the annual equivalent of 3.6% of GDP. The Fed’s recent raise of 25 bps in the fed funds rate and ongoing higher oil prices is unlikely to be sufficient to slow the AI trade given the magnitude of the spending. The demand inflection for AI has pushed hyperscaler capital expenditures to a projected $781bn in 2026, up 88% year over year (P. 26) and to 39% of revenue (and $1.07 trillion estimated for 2027, 44% of revenue), and GPU contract rental pricing up roughly 65% YTD, with on-demand capacity largely unavailable since February. Compute scarcity is driving up the cost of serving AI even as token prices fall, down 41% since March, and token economics is becoming a bigger issue. This is demonstrated by Anthropic's push toward metered APIs and usage- based pricing, which has driven its meteoric revenue growth over the past 12-18 months, to $11.5bn in Q2 2026, or about $46bn annualized and increasing enterprise interest in lower cost open-source models.

The exit environment remains narrow, and the IPO market has become more subdued. Excluding SpaceX's acquisition of xAI and its IPO, US tech exit volume reached $219bn in the first three quarters of 2026, already above all of 2025, but a single deal, SpaceX's $60bn acquisition of Cursor, accounted for 75% of Q3 volume. Q3 had 28 exits, the fewest in three years, and strategic M&A made up 80% of exits by count this year. The wave of listings expected to follow SpaceX has slowed. Anthropic is reported to be looking at a November pricing and OpenAI has delayed their IPO into 2027.

Valuations increasingly depend on which layer a company sits in. Our Rule of 40 regression explains 41% of the variance in revenue multiples. Cybersecurity names trade almost entirely above the regression line (premium valuations) relative to growth and margins and application names mostly below (discounted valuations).