Here’s something that didn’t make many headlines: in May 2026, the iShares Expanded Tech-Software Sector ETF (IGV) returned 21.2% on a month-to-date basis, outpacing the Semiconductor ETF (SMH), which returned 19.8% over the same stretch. The same group that Wall Street had essentially declared dead three months earlier just beat the most-hyped trade in the market.
Let that sit for a second.
The context matters here. Earlier this year, the software sector got absolutely punished. The fear was straightforward and, on the surface, convincing: if AI agents can write code, resolve customer tickets, and automate workflows autonomously, why would enterprises keep paying expensive per-seat licensing fees? The “SaaSpocalypse” was the name Wall Street gave it. By April 2026, the selloff had wiped roughly $2 trillion from software stocks, with names like Salesforce, ServiceNow, and Snowflake all losing more than half their market cap at the trough. Software forward price-to-earnings multiples, which once traded at 84x during the 2020–2022 peak, had collapsed to just 22.7x by March 2026 — briefly falling below the S&P 500’s overall market multiple for the first time in the sector’s modern history.
That was the dislocation. The question now is whether the recovery is real or a dead-cat bounce.
The Earnings Tell a Different Story
What changed the conversation wasn’t sentiment — it was results. Twilio reported Q1 2026 revenue growth of 20% year over year, raised its full-year guidance to 14–15% top-line growth and $1.08–$1.10 billion in free cash flow, and added 43,000 net new accounts in a single quarter — more than the company’s entire full-year net adds in 2023. MongoDB delivered a second consecutive quarter of accelerating revenue, with Atlas growing 29% and the highest first-half net new customer additions the company has ever posted. Snowflake ripped 49% in a single week after its AI-driven results came in well ahead of expectations.
Cloudflare has now strung together 11 consecutive quarters of 27%+ growth. Its FY2026 guidance calls for 28–29% revenue expansion. Net revenue retention at Snowflake came in at 125%. These are not the metrics of a sector in structural decline.
Slight tangent, but it matters: JP Morgan published a research note earlier this year arguing the software selloff had gone too far — that fears around AI disruption were "overblown" and based on "broken logic." Their argument centered on long-term contracts and high switching costs that create durable moats competitors can’t circumvent overnight. Morgan Stanley’s head of U.S. software research said the same, flagging Microsoft, Salesforce, ServiceNow, Snowflake, Cloudflare, and Palo Alto as underpriced given their actual product cycles. Jefferies noted that nearly 75% of U.S. software stocks were screening as technically oversold — the highest reading in the firm’s recorded history.
The market didn’t listen. Then earnings season showed up and made the case louder than any note could.
The Bifurcation Nobody Is Mapping
This is where it gets interesting. The recovery isn’t uniform. The SaaS sector has quietly split into three distinct categories, and most investors are treating it like one.
- AI infrastructure software (Cloudflare, Twilio, Snowflake): These are the picks-and-shovels of the agent era. AI agents need somebody’s network layer, somebody’s data warehouse, somebody’s voice API. These companies are the invisible backbone. Re-accelerating.
- AI-attached applications (HubSpot, Atlassian): Legacy subscription models being rebuilt around AI as an expansion vector. HubSpot has held 19–21% revenue growth while integrating AI features. Atlassian now has 600+ customers above $1 million in ARR, up 40% year over year. Growth is holding if the AI layer adds pricing power.
- Seat-model incumbents under pressure: Companies where the disruption risk is real — pure workflow automation plays without deep data moats or infrastructure positioning. These are the ones that may not recover at the same velocity.
Gartner projects enterprise software spending at $1.43 trillion in 2026, with AI agents embedded in 40% of applications. Global software spending growth is forecast at 14.7% this year. The "AI will kill SaaS" thesis assumed a zero-sum game. What’s actually emerging looks more like a platform shift — where infrastructure-layer software wins and the weakest seat-model vendors consolidate or fail.
What the Technical Picture Shows
IGV bottomed in early April and has since staged a 44% rally off those lows, bringing its year-to-date performance back into positive territory for the first time. The fund is now less than 9% below its all-time high from September 2025. Cybersecurity names have led the charge — the Amplify Cybersecurity ETF (HACK) is up more than 30% year-to-date, driven by CrowdStrike (+67%) and Palo Alto Networks (+63%).
Key resistance sits at the September 2025 highs. A clean break above that level on volume would confirm a new leadership phase, not just a relief rally.
Bull / Base / Bear
- Bull: AI infrastructure demand drives a supercycle in data, networking, and automation spend. Snowflake, Cloudflare, and Twilio emerge as the platform winners. IGV reclaims all-time highs by Q4 2026. Enterprise AI adoption reaching scale triggers a new round of SaaS spending.
- Base: The bifurcation continues. Infrastructure names extend gains. Seat-model incumbents stabilize but don’t re-rate meaningfully. IGV trades in a range near current levels as investors pick through the wreckage selectively.
- Bear: AI agents accelerate faster than incumbents can adapt. Enterprise IT budgets consolidate into fewer vendors. Per-seat revenue compression hits even the strongest names. The recovery stalls and the sector revisits April lows.
The base case is probably the most intellectually honest — but the bull case has more evidence behind it right now than most analysts expected three months ago.
What investors are likely underestimating: the infrastructure layer of the AI stack is software. Cloudflare’s network, Snowflake’s data warehouse, Twilio’s communication APIs. The "AI will kill software" trade was based on a misread. The real risk was always which software — not software itself. That distinction is now starting to matter in price.
For informational purposes only.
