Eight months ago the market decided that AI was going to eat B2B software. Roughly $2 trillion of market cap came out of the group. Then it came back.
The recovery at the index level is close to complete. The iShares Expanded Tech-Software ETF (IGV) closed September 3 at $106.81, up about 40% from its April low near $74. It gained 16% in August. State Street’s XSW software ETF hit an all-time high in late August. Jefferies told clients over the Labor Day weekend that they expect the momentum to run into September and that AI displacement fears are overblown.
Inside that recovery sits a 117-point spread. CrowdStrike is up 83% on the year. monday.com is down 36%. Both sell software to B2B buyers, both had relatively strong growth, albeit a different slope to that growth.
Five things to take from the year so far:
- The index recovered and the constituents did not move together. CrowdStrike, Twilio, and Snowflake are all up more than 70%. HubSpot and monday.com are down more than 33%.
- Growth rate explains some of the tape and nowhere near all of it. Twilio and monday.com both grew 22% in their June quarters and finished 109 points apart.
- Two other variables did most of the sorting: whether AI workloads land on your meter, and what multiple you carried into January.
- The big incumbents took the largest drawdowns and made most of it back in a single month. Salesforce set a 52-week low of $146 on June 22 and finished September 3 up 4.3% on the year.
- The private market kept score too. Airtable, worth nearly $12 billion at its 2021 peak, sold to Bending Spoons for under $1.3 billion.
The winners: consumption and security, up 54% to 83%
- CrowdStrike: up 83.4% at the September 3 close. ARR of $5.84 billion growing 25%, with net new ARR growing 51%. Falcon Flex ARR passed $2.29 billion, up 101%, and is now roughly 39% of total ARR.
- Twilio: up 73.8% at the September 3 close, $240.48 from a $138.36 start. Q2 revenue of $1.50 billion grew 22% reported and 17% organic, and the full-year guide went to 18% to 18.5% from 14% to 15%. Worth reading the fine print: $71 million of the gap between reported and organic growth is carrier fees on business text messages, which Twilio passes through at cost.
- Snowflake: up 73.1% through the September 2 close, and that figure understates it. Product revenue of $1.49 billion in Q2 FY27 grew 37%, the third straight quarter of accelerating growth, and the stock jumped sharply on September 3 after the report. Guidance raised to $6.07 billion.
- Datadog: up 53.9% through the September 2 close. Q2 2026 revenue grew 35.6%. More than 750 AI customers on the platform, including all ten of the largest AI companies.
Every name in this group either bills by consumption or sells into a budget that AI expands. When an enterprise deploys agents, it generates more logs to monitor, more data to query, more traffic to route, and a larger attack surface to defend. Datadog gets paid whether the telemetry comes from a human or an agent. Snowflake gets paid regardless of who runs the query.
Security is the cleanest version of the argument. George Kurtz credited CrowdStrike’s momentum to the world’s adoption of AI rapidly expanding the attack surface. The disruptor is the demand driver.
One caveat: Datadog fell 19% on August 6, its steepest drop since its 2019 IPO, on a quarter that beat and raised. Management disclosed that its largest customer was reducing usage. Consumption pricing gives you the upside on AI workloads and hands you customer concentration risk in the same sentence.
The round trip: down 33% to 57%, then most of it back in August
Salesforce was down 29.5% year to date in early April, to +2%. It ran it all back. It set a 52-week low of $146 on June 22, more than 30% off its late-December high. It reported Q2 FY27 on August 26 and gained about 40% over the course of August. It closed September 3 at $264.43, up 4.3% for the year from a $253.62 start. Underneath: cRPO grew 14% in constant currency, while organic revenue grew 6.4% once you strip out Informatica’s $456 million.
Atlassian ran the widest range in the group, from -7% to 24% over the year!. Down roughly 57% at the March lows, with a 52-week low of $56.01. It reported Q4 FY26 on August 6, jumped 35.31% the next day, gained 88% over the month of August, and closed September 3 at $194.68, near a 52-week high of $198.56. That leaves it up 20.1% on the year. Q4 revenue grew 28%, cloud grew 31%, and RPO grew 44%.
Look at what the recovery was priced on. Salesforce disclosed $1.5 billion of Agentforce ARR and Slackbot passing 1 million users in five months. Atlassian disclosed more than 1 million monthly active users on its MCP server and Teamwork Graph CLI. In each case the re-rating followed evidence that AI adds a revenue line, and it arrived months after the growth rates had already been reported.
The losers: seat-priced platforms, all down more than 33%
HubSpot is down 33.3% year to date, at $255.12 on September 3 against a $382.23 start. On August 6 it had its worst day on the market in a decade, falling 19.47%. The Q2 numbers underneath are respectable. Revenue grew 20% as reported and 17% in constant currency. But net customer additions came in at 7,000 against a 9,000 to 10,000 expectation, net revenue retention slipped a point to 102%, customer dollar retention sat in the high 80s, and Yamini Rangan’s team guided quarterly net adds down to 5,000 to 6,000 for Q3 and beyond. The full-year guide came down. Management named the cause on the call: larger buying committees, C-suite and board approvals, and longer sales cycles as budget scrutiny increased.
monday.com is down 35.5% through the September 2 close. Revenue grew 24% in Q1 and 22% in Q2. The full-year guide is 18% to 19%, down from a Q1 raise. The company restructured in Q2 and moved to consumption-based pricing for its AI products, and AI ARR doubled quarter over quarter to 17% of net new ARR. The market gave it nothing for that.
Figma is down 35.1%, at $24.40 midday September 4 against a $37.61 start, on a $13.0 billion market cap and a 52-week range of $16.60 to $71.48. Figma is the fastest grower in this entire post at 48%, with 136% net revenue retention, and it is the third-worst stock in it. The easy explanation is that this is the July 2025 IPO pop unwinding, and that explanation is wrong: Figma entered 2026 at $37.61, already far below the roughly $47 billion it closed at on day one. The unwind happened in 2025. This is a clean 2026 decline on top of it.
Two things did it. Figma reported a beat-and-raise in August with three straight quarters of accelerating growth and fell 15% anyway, because the Q3 guide implied about 1% sequential revenue growth. And gross margin came down five points year over year as AI credits added a real inference line to COGS. Figma is seat-priced, per editor, and it is paying for AI on the cost side while the pricing unit stays where it was.
The private market kept score in the same direction. Airtable, valued at nearly $12 billion at its 2021 peak, agreed to sell to Bending Spoons for under $1.3 billion. Bending Spoons, which went public in July, now carries a market cap around $25 billion.
Some of this group is now being priced as takeout material. Reuters reported that Workday is in acquisition talks with Silver Lake. Workday grew total revenue 12.8% in Q2 FY27 and subscription revenue 13.9%, with 12-month subscription backlog up 14.2% and total backlog up 8.0%.
Twilio and monday.com both grew 22%. One is up 74%, the other is down 36%.
Twilio and monday.com both grew 22% in their June quarters and finished 109 points apart. HubSpot grew 20% and lost a third of its value. CrowdStrike grew ARR 25%, five points faster than HubSpot, and gained 83%. Salesforce grew 6% organic and is up on the year, while HubSpot grew more than three times faster and is down.
Growth still gets paid. Snowflake and Datadog are growing in the mid-30s and both had strong years. But growth by itself would have gotten you the two ends of the tape backwards.
Two other variables did most of the sorting.
The billing unit. Every name in the winners group prices by consumption, by commitment pool, or by protected surface. Every name in the losers group prices by seat. When a CIO reallocates budget toward tokens, the seat line is the line that gets optimized. Zylo’s 2026 index found average SaaS spend of $55.7 million on a flat 305-app portfolio, with 79% of companies hitting a renewal increase and 61% cutting projects. HubSpot’s commentary about budget optimization headwinds and longer approval chains is that trend showing up in a P&L.
The multiple you carried into January. Salesforce entered 2026 at a forward P/E of 14.1 with a PEG of 1.0. When the panic reversed, a good quarter moved it 40% in a month. The names that entered the year priced for perfection had no such room, regardless of how fast they were growing.
For founders benchmarking against public comps, that is the number to internalize. In 2026 the public market paid for revenue that moves in the same direction as AI adoption, bought at a price that had already been beaten down.
What Happens The Next 6-12 Months?
Do the consumption names hold their gains through a full budget cycle? The bull case for Snowflake and Datadog has one quarter of AI-driven results behind it. Datadog’s largest customer cutting usage is a reminder that a meter runs both directions. The next print is the test.
Does HubSpot’s net adds stabilize at 5,000 to 6,000? That guide is the single most important number in mid-market B2B right now. If net adds hold there and net revenue retention stops sliding, the drawdown was an overreaction. If both keep drifting, the market got there first.
Do the markets re-rate on a consumption-model model change? monday.com moved to consumption pricing for AI and doubled AI ARR in a quarter, and the stock did nothing. Salesforce is running four pricing models at once, and Marc Benioff said on the Q2 call that the company is still trapped in some ways in old per-user pricing models. Until the market pays for one of these transitions in advance, every incumbent takes the drawdown first and the credit later.
Do the take-private bids keeps building?. The Workday and Silver Lake report is the loudest version of it, and the Airtable outcome is the same signal from the private side. A large cohort of 12% to 20% growers with real cash flow and a compressed multiple is what PE buys.
The 2026 spread between the best and worst performer in this group is 117 points. The variable that explains most of it is what unit sits on the invoice, and whether that unit goes up when your customer deploys more agents.
Year-to-date figures are through the September 3, 2026 close for CrowdStrike, Twilio, Atlassian, Salesforce and HubSpot, and through the September 2 close for Snowflake, Datadog and monday.com. Sources are Google Finance for Twilio, Salesforce and HubSpot, and MarketBeat for the rest. Growth rates come from each company’s most recently reported quarter.
