TechStockFundamentals

I think there are basically 2 prevailing strategies in the market. 1) Earnings revisions Only two questions matter near-term: Is the story getting better or worse, and are numbers going up or down? That's it. For momentum names, in either direction, that's all that matters -> is the story continuing to improve w/ beat and raises, or is the story continuing to erode with misses / guidance reductions. There is more complexity at extremes (ie second derivatives) that also works -> the story has stopped getting worse and numbers are no longer going down so upside optionality, or narrative no longer improving and downside risk to numbers so downside optionality. Earnings revisions work because they are reflexive. The company is doing well, estimates are moving higher and the stock is rising. More investors do work on the name and the due diligence will be positively biased as they generally ask more favorable questions to understand the strengths of the company. This leads to more buying and the increased stock price reinforces the bullish narrative. Hopefully, management is smart enough to set forward guidance at beatable levels and the step-ups continue. Same is true for reflexivity on the downside. A company is struggling and missing its own guidance and expectations. The stock moves lower as obviously more people sell than buy. Investors do more work but are naturally more focused on the reasons for corporate underperformance and corresponding risks. Further stock price declines fuel the negative narrative which feeds on itself. Companies often don't take their medicine as far as fully disclosing bad news and the negative revisions continue. So again, the playbook for this is simply 1) is the story getting better or worse?, and 2) are numbers going up or down? 2) Valuation From time to time, stocks get too cheap or expensive relative to fundamentals such that the disconnect is very interesting (let's say can make at least 50% to get back to conservative valuation target in either direction within 12-18 months). Very often, large disconnects occur when earnings revisions are unclear to everyone (stock is really cheap but no one has any clue when numbers stabilize, stock is really expensive but everyone thinks company will continue beating / raising). However, the insight here is that valuation is interesting in and of itself for the right kinds of companies (solid business, transitory issues, acquisition candidates, etc.). And that no one has visibility into near-term numbers is already in the stock and why the valuation is so compelling. This opportunity also always arises in tough macro periods where everyone is frozen bc who can predict how the great financial crisis or covid pandemic plays out - so that uncertainty is shared by everyone and more than represented in stock valuations. And that is why the opportunity exists - because everyone is worried about the same imponderable thing and it's already in the stock price. The only advantage you need here is a deep understanding of and conviction in the business, and freedom / willingness to buy it for the long-term (and the returns never take as along as you think). Reality My experience is most investors typically only pursue one of above strategies (and might only be allowed to pursue one or the other). So they won't / can't buy $WDAY at 20x FCF bc they think guidance might be lowered (company misexecution, tough macro, etc.), or they won't sell $WDAY at 35x FCF bc they are confident everything is great and company will continue beating. I also have found earnings revisions investors tend to have nearer-term horizons and be less valuation sensitive. This is because it's hard to fill a portfolio with names that will both beat earnings and trade at reasonable valuations. So, on balance, these investors will compromise on valuation. Valuation investors have longer-term horizons and are by definition valuation sensitive and less sensitive to revisions. This is because it's hard to fill a portfolio with names that are both cheap and that will beat earnings. So, on balance, these investors will compromise on earnings visibility. Optimal I have found earnings revisions offer the most explanatory power near- to mid-term. I want to be long stocks where numbers are moving higher and short stocks where revisions will be lower. But I also want think it is an advantage to be able to cast that strategy aside and buy companies at valuation extremes. This is when a good company is "too cheap" even if I'm not sure of near-term revisions. There are other criteria that need to be in place for this to work -> solid business w/ reasonable estimation capabilities, potential negative revisions are manageable and likely in the stock, less risk of permanent impairment over a 12-18 month horizon, balance sheet in order and not indebted, etc. Finally, the intersection of the two is where you can nail inflection points in businesses / stocks. A company is too cheap and is about to start beating numbers which will drive both estimates and the multiple higher. That's where the big money resides. Anyway, this overall thinking is why I structure my notes as follows: Is this a bad, good, or great business? Is the story getting better or worse? Are numbers going up or down? What is the risk / reward?

@abshah101 I'm probably >100 bps higher of PF OM and was understating amortization build up from acquisitions.

@TechFundies Company is guiding to ~11% GAAP margin this year. You are modeling 16.8%? What am I/the market missing? I see lower SBC and amortization adding 8-10% by 28/29, so low 20s GAAP margin vs you at 28%. Seems optimistic?

$NOW analyst day was solid. This is the operational workhorse for large enterprise workflows 1) expanding its footprint across functional division most recently into security / CRM, 2) embedding AI across the platform which transitions customer relationships to actions and usage pricing, and 3) moving down market on margin. They are firm that they can’t / won’t open up the platform entirely but are going to construct autonomous work units that can be triggered outside the system but will be monetized [boo hoo AI bears. You don’t get to go to Disneyworld with your own food and operate the rides yourself either]. The financial guidance is a little head-scratching to me. The 2030 subscription guide of 30b base case and 32 plus case implies subscription growth steadies around 20% which would certainly be unique and feels optimistic. Fine. But, then the margin guidance for 100 bps expansion in 2027 feels way too conservative as they said 1) subscription GM will not dip below 80% (AI costs are negligible), and 2) employee count will be flat yy despite the acquisitions. So the exit margin in Q4 is going to be monstrous and run-rate into 2027. Said differently the margin guidance feels inexplicable conservative. SBC will be <10% of rev in 2029 so very nice progression there as well. Whatever, stock remains stupid cheap here. Numbers in my model include amortization in GAAP which is 5% of profit in 2028. Notes AI -Birthrates around the world going down and headcount at corporations flat. -AI <100 day guarantee on AI go-lives to predictable ROI -Re-wrote AI stack to be built-in. ->700 customers have consumed >500m credits -Not going to just open up NOW to anyone but going to open up system of action – open up workflows, skills, context engine, etc. – and will monetize it -Monetization -Will just monetize any action taken inside the system regardless of whether by person or agent Products -Security / Risk – think can be much larger bc didn’t try that hard to get it here. Expanding into identity governance. -Crossed 1b in ACV by 3Q25 w/ NNACV growing 40% yy -<20% penetrated in install base -Business continuity mgmt. and integrated risk mgmt. -Armis / IoT security -Veza is an access graph to understand who / what is accessing what and then build apps on top -Data / Analytics on track to be 1b - >2x NNACV yy growth -RaporDB did 100m in first full year ACV -CRM will be at 2b soon – sales automation, CPQ, Field service mgmt., customer service mgmt -25m employees in employee center Sales -Have 9/10 F500 so going to take down market a bit in upper / mid-market Financials -Will exit this year w/ headcount that they had entering this year -Rule of 60+ by FY30 and >30b of subscription revs – “Whole mgmt. team bought into this from a bottoms-up basis” -SBC <10% in 2029; rule of 50+ GAAP -Expect 20-30% price uplift from new price packaging which introduces consumption earlier -Customers who are existing Now Assist expanded ACV by 3x due to buying more licenses and AI

@aenigme What’s your evidence for any of this?

@TechFundies Servicenow is a deadman walking. "Have 9/10 F500" - Their moat has been destroyed, they are slashing prices and there is no one left to upsell to. "embedding AI across the platform" - Glorified chatbots and their SDK has zero access to the platform scope (and never will).

Random thoughts on $MNTN from the weekend: 1) The gap between META ad quality and CTV is miles wide. Watched some HBO Max ad-supported tier over the weekend and was just body-slammed with ads for Euphoria, Folger's coffee and Ro repeatedly. Why?! Kind of wild how poorly that very valuable inventory is being treated. Just tons of win-win by bringing in long-tail advertisers and matching them effectively. ($MNTN focuses on long-tail). 2) As a side project, I'm helping someone build a business and, for fun, running some of their digital ads. The META ad tools are laughably bad. Despite their purportedly being a real player in models, there aren't even obvious integrations in the ad-buying process. For instance, there is no ability to generate pics for an ad type, generate tagline suggestions, or generate videos from images. You have to pop over to ChatGPT, Gemini, Runway, etc. to get any of that done! I went over to QuickFrame AI (the AI video-gen site owned by $MNTN) and tried the new SeedDance video generator using just a few still images. It was pretty good on first pass. Not TV quality but I'm sure other people can use the tool to get it to that level. Just seems straightforward that AI tools are going to enable the long-tail of the market to create sufficiently compelling video ads to more easily participate in CTV, and that CTV ad inventory needs long-tail participation. All positive for $MNTN where the story is getting better, numbers seem pretty low and stock is dirt cheap. The risk / reward on this one is almost too good to be true - which means either I am or the market is very, very wrong. Could be me... Let's see what we get with earnings tomorrow. Q1 is not exactly a barn-burning quarter. But worth noting all software / Internet stocks are pretty beaten up and maybe that starts turning at some point this year.

Gemini's view of my post: The idea of a joint conference call is hilarious—we’d finally get to see Satya, Andy, and Sundar dodge the "circular revenue" question in real-time while comparing who has the bigger power grid. It would certainly save everyone from the 12-hour earnings-day marathon. Your spreadsheet captures the "infrastructure-only" view much more cleanly than the sell-side reports that often muddle AMZN’s logistics spend with its data centers. That normalization is key to seeing the true ROIC of this cycle. Here are a few thoughts on the numbers you’ve laid out: 1. The Productivity Step-Down ($0.42 \rightarrow 0.31$) Your assumption of a meaningful step-down in productivity is probably the most realistic part of this model, yet it's the one the market is currently trying to ignore. That drop from $0.42$ to $0.31$ represents the "Inference Tax." The Mix Shift: Earlier CAPEX was heavily weighted toward training (lower volume, higher intensity). As we move to $2026/2027$, more of that spend goes toward inference capacity. The ROIC Compression: Even if demand stays high, the competition for tokens is driving prices down faster than the hardware is getting cheaper. A $0.31$ productivity ratio might actually be generous if we see a price war in frontier model APIs. 2. The Model Company "Revenue Gap" You noted Anthropic and OpenAI adding $\approx \$60\text{B}$ in ARR. That’s incredible, but your 2027 revenue projection requires an incremental $\$161\text{B}$ in cloud revenue. The Delta: Even if the model companies spend $100\%$ of their ARR on compute, there is still a $\approx \$100\text{B}$ "gap" that needs to be filled by enterprise SaaS integration (the Fortune 500 actually moving from "POC" to "Production"). The Risk: If the "Agentic AI" wave doesn't start producing massive enterprise billings by mid-2027, that $0.31$ productivity number starts looking like a ceiling rather than a floor. 3. The Law of Large Numbers & The "Second Derivative" Your question about the second derivative turning negative in 2027 is the "sweaty palms" moment for the semis. Semi Visibility: Broadcom and NVIDIA are already touting visibility into 2028, but as you know, "visibility" is often just another word for "non-cancelable backlogs" that can become very painful if the ROIC doesn't materialize. The Peak: If CAPEX growth slows to $28\%$ in 2027 and flattens in 2028 (per your table), the "growth" narrative for the hardware providers might peak 6–9 months before that, which puts the danger zone right in early 2027. 4. Valuation: "Cheap on This Year" You’re right that these stocks look like a steal if you only look at 2025/2026 growth. The "HYPE" complex is essentially a bet that the model companies' revenue growth is a leading indicator for the broader economy’s AI adoption. As long as OpenAI and Anthropic are "scorching," the market will likely look past the CAPEX-to-revenue ratios (which are hitting $40\text{--}50\%$ for MSFT and GOOGL). The moment an OpenAI revenue target is missed (like the rumors we saw last week), the "productivity" column in your sheet becomes the only thing investors will care about. One follow-up on your "Invert it" approach: When you look at that 2026 CAPEX plan of $\$513\text{B}$, how much of that do you think is "defensive" spend (building just because they can't afford not to have the capacity) versus "offensive" (actual line-of-sight to 2027 revenue)?

ChatGPT's view of my post: This is a strong framing. You’re doing the right thing by forcing everything through capacity → revenue → productivity, rather than getting lost in anecdotes about demand. A few reactions, pushing where I think it matters: 1) The core insight is right: 2025 validates, 2026 is the bet 2024 → 2025: productivity improves (0.39 → 0.42) despite scale → that’s the key bullish data point That tells you this isn’t just overbuild (yet). Utilization + pricing + mix are holding. So I agree: 2025 capex now looks “earned”, not speculative. But… 2026 is where you’re explicitly underwriting a productivity decline (0.42 → 0.31) That’s doing a lot of work in your model to keep things sane You’re basically saying: “Things get worse… but not too much worse” That’s reasonable—but also where error bars explode. 2) The uncomfortable math: absolute dollars now dominate everything You highlighted it, but it’s worth making sharper: Incremental rev: 2026: +$116B 2027: +$161B 2028: +$199B At this point, the system needs to “find” a Fortune 50 company worth of new revenue every year just to keep ratios intact. That shifts the debate from: “Is demand strong?” → clearly yes to: “Is demand deep enough across many buyers?” Because concentration risk becomes the whole game. 3) Your biggest hidden assumption: model company reinvestment rates This is the most important line in your writeup: assume model companies spend ~100% of ARR on infra That’s doing a ton of heavy lifting. Three things to pressure-test: Training vs inference mix Training is lumpy and front-loaded Inference is recurring but margin-sensitive Vertical integration risk If OpenAI / Anthropic internalize more stack over time, hyperscaler “tax” compresses Unit economics inflection If inference gets dramatically cheaper (hardware + model efficiency), revenue can grow while infra intensity falls That last one is the real bear case—not demand collapse, but decoupling. 4) Backlog normalization: good instinct, but fragile signal Your “duration-adjusted backlog” is directionally right, but: Longer contracts ≠ guaranteed utilization Pre-buying capacity in a supply-constrained environment can inflate backlog without true end-demand visibility So: The 77% YoY growth is real signal The absolute 350B number is less trustworthy as a denominator 5) Where I’d sharpen the framework Instead of a single productivity number, I’d break it into: Productivity = (Utilization) × (Price) × (Workload mix) Right now: Utilization: very high (tight supply) Price: stable to up (AI premium) Mix: shifting toward inference (good for durability, maybe worse for margins) Your 0.31 assumption implicitly bakes in deterioration across these—but it matters which lever breaks. 6) The real “tell” isn’t 2026—it’s early 2027 signals You said it well: 2026 probably works. The market will key off leading indicators like: Growth rate of inference tokens / usage Signs of price erosion (per token / per GPU hour) Whether new buyers emerge beyond the current few mega-model companies If those broaden → your 2027 numbers are achievable If not → productivity likely undershoots your already conservative 0.31 7) On the stock call This is the cleanest version of your thesis: “While model companies are scorching, everything works.” I’d just tighten it one notch further: Near-term (next 2–3 quarters): Setup is strong, numbers likely beat, hard to fight Medium-term (’27): You are long reinvestment intensity, not just demand That’s a subtle but important distinction: Demand can stay strong Stocks can still struggle if ROI narrative cracks Bottom line Your framework is one of the better ways to attack this 2025 data supports the bull case more than skeptics expected But the entire debate is now shifting to: Can the ecosystem sustain capital intensity at this scale—not just demand? That’s a much harder question than “is AI growing?” If I were to stress test one variable in your model, it wouldn’t be growth—it would be the 0.31 productivity assumption. That’s the fulcrum where this either all works… or starts to unravel.

Since all the hyperscalers are going to report on the same day, maybe they can just hold a joint conference call to make it easier on us? Would make Q&A a lot more interesting. Ok – so the numbers were so wild that I think we have to start by looking at them together : $AMZN AWS + $MSFT Azure + $GOOGL GCP = HYPE. HYPE cloud rev of 84.4b grew 6.8% qq / 39.1% yy with net add of 5.4b. Just to provide some context for how outrageous this is, the 5.4b in net add compares to 0.7b last Q1 and 1.4b the Q1 prior to that. (FX was a tailwind of a few points but inconsequential as far as the bigger picture). I attempt to normalize backlog for duration (ie rough exercise to normalize for companies signing longer-term deals) which results in “current” backlog of 350b up 77% yy. Net add add of 68b in Q1 up from 19b last Q1 and 5b the Q1 prior to that. Outrageous. GAAP margins were healthy across the board which validates capex spend to date. Digesting this AI infra cycle is really mind-bending so let’s just try to invert it all and start with facts. I try to normalize all CAPEX for HYPE infrastructure only (and strip out capex for AMZN retail, GOOGL search, etc.). Here is HYPE CAPEX by year relative to beginning current backlog -2024: CAPEX of 153b was 97% of starting current backlog -2025: CAPEX of 274b was 154% of starting current backlog -2026: CAPEX plan of 513b is 182% of starting current backlog. And is 147% of current backlog at end of Q1 (might be more fair given massive Q1 bump). 2025 / 2026 CAPEX are clearly elevated versus starting backlog compared to 2024, but also include more longer-dated investments as a percent of total. Also demand is exploding so leaning in makes sense. So if 2025 CAPEX was rational based on the revenue ramp / healthy margins we are seeing thus far, 2026 also seems defensible relative to starting backlog. In other words, the 2025 CAPEX bet is playing out nicely and 2026 seems inline with 2025 based on backlog. Here is capex revenue productivity (incremental cloud revenue in forward year / current year CAPEX) -2024 productivity of 0.39 (2025 incremental revenue / 2024 capex) -2025 productivity of 0.42 (2026 estimated incremental revenue / 2025 capex). This is based on my estimate that HYPE does 393b in rev this year which is up 42% yy. So what will revenue productivity be in 2026 (ie how much incremental revenue in 2027 from capex this year?). I am assuming a meaningful stepdown in productivity (competition for ROIC, more long-dated investments, etc.) to 0.31 down from 0.42 in 2025. This implies HYPE will do 554b of rev in 2027 up 41% yy w/ net add of 161b up from 116b this year. So, we are now crossing over from facts to implication. 2026 is probably going to be great. (Or maybe this massive Q1 / Q2 spending ramp results in CFO panic / scrutiny and we see an equally epic sequential slowdown in Q3 / Q4. Don’t know.) Anthropic / Open.AI have added 50-60b in incremental ARR thus far this year (!). Let's say they add >100b. An overly conservative estimate would be that they spend 100% of this ARR on HYPE (+ORCL / NEO) - think of it as 50% of the spend is for inference and 50% for forward model training (we know model training is way higher). Anyways, 2025 capex seems like a solid investment w/ cooked 2026 revenue growth. So the trillion dollar question is will the model companies (Open.AI, Anthropic, Gemini, ...) see scorching revenue momentum in 2027 to justify 2026 CAPEX and justify meaningfully higher CAPEX in 2027 and maybe 2028? These feel like the tip of the spear questions to me: -Can the HYPE capex of 513b this year generate 161b in incremental rev in 2027 (up from 116b incremental in 2026)? Going to need continued, MASSIVE growth out of the model companies... -If this year goes according to plan, will HYPE capex for 2027 reach 658b (28% growth)? On the back of that, can HYPE 2028 revenue grow 36% growth at 0.30 productivity - this implies 199b in additional revenue dollars?! I don't really know how to put odds on that. That's a lot of incremental revenue dollars. I’m not even including ORCL and neoclouds in this capex / revenue capacity / productivity exercise which would make hand-wringing even sweatier (tons of capacity coming from them which makes the revenue bogey to preserve ROI even higher, or completely whacks industry ROI). I can’t really answer these questions with confidence. That raises unanswerable follow-on questions – does the law of large numbers dictate capex has to peak in or by 2028? And does that imply second derivative turns negative by 2027? (if so, semis probably peak later this year)? Is HYPE cheap here ahead of great growth / ROIC? Business this year seems locked in given model momentum and the stocks are cheap on this year. So arguably even if model momentum slows meaningfully in 2027, we are already pricing in lower ROIC on 2026 capex. Maybe the simplest view ends up being the correct one: while the model companies are scorching, the stocks will do well. And whenever / however that party ends, the whole ecosystem goes down. Thoughts welcome but only if superior to AI thoughts below which are pretty strong. If AGI is measured relative to average comment on X or qualitative feedback from the average investment team, we're already there. Seriously.

One market behavior that seems consistent over time is shorts piling in at bottoms. Who hasn't done it from time to time (I'm guilty of it from my hedge fund days)? Just don't let the stock drawdown trap you in your false narratives on the way up:)

I think the "easiest" names are the platform enterprise SaaS businesses that are vital solution providers to their customers that are customized to vertical needs and provide cross-departmental products that enhance stickiness and offer the greatest opportunity to make AI additive to the customer value-add and financial relationship. $NOW, $CRM, $WDAY, $MSFT, $SAP. $TEAM likely also fits this description though closer to the front lines of this battle.

@TechFundies Thanks for the update! Would love to know what handful of names you think are most durable and will ultimately be a big ai winner. Theres bound to be some and the asymmetry is exquisite!

$TEAM reported a fine qtr. Cloud rev accelerated to 29% yy though probably closer to low 20s normalized for migrations. The punchline is their users are growing, usage is growing, more modules being adopted partly through bundling, dollar net retention improving, etc. This one is still in the thicket of confusion since it is so close to AI code development, but the truth is likely AI development cycles involve 1) design what needs to be done and put it into Jira, 2) take that ticket, 3) download codebase from Git, 4) use Codex / Cursor / etc. to suggest code for components, 5) test it, 6) check-it back in, 7) complete documentation. You and your buddy can vibe-code your dog-washing app but Bank of America doesn't really operate that way. GAAP OM was 10% normalized for restructuring charges. Founder-CEO probably gets it - bring up GAAP margins even while investing in AI, stop buying browser companies, buy-back stock (they bought back 4% of outstanding in the qtr). I'd do more parsing of corporate detail but it doesn't really matter. Either it's dead because of how AI development changes over a few years or it's going up 2x just back to a non-dead multiple and probably 3-4x over 3-4 years. Even w/ the big after-hours move, stock trades at 11x forward FCF. Worth noting with SaaS stocks down >50%, that SBC running through the P&L is probably overstating actual cost by >50% too. Notes AI -Can assign tasks to Rovo and other agents in Jira – iterate w/ them and embed agents into workflows -Agent activity captured alongside task history -Remix converts Confluence into images -Rovo in Service -Rovo customers growing ARR at 2x non-ROVO -AI credit usage growing 20% mm -Teamwork customers use 2x more AI credits / paid user Sales ->55k cust >10k in cloud ARR ->600 customers with >1m ARR -350k total customers -Not seeing any evidence of seat compression – actually seeing seat expansion in core Jira customers as well as beyond -DBNR ticked up again Service Collection -Jira + CSM + Assets + Rovo -Crossed 1b in ARR and growing >30% yy ->60% of instances are for non-IT functions -Have customers that have 500 service desks across their organizations ITSM -Largest competitive displacement qtr -Human supervised AI agents plans and executes employee onboarding and support. -AIOps is AI-driven early incident detection and change risk assessment. See >70% AI resolution rate across more than 100k conversations. Qtr -Rev up 32% yy -Cloud accelerated to 29% yy – seat expansion in Jira and adoption of Teamwork for AI capabilities [Also conversions from on-prem] -Pricing change in March as well -More upfront term recognized - 50m more than expected -DC grew 44% due ot end of life acceleration -RPO of 4b up 37% yy; cRPO grew 22% yy (end of life pushed more to term recognized up front) -cRPO would have grown >30% normalized -PF OM of 34% up from 26% last Q3 -10% GAAP OM excluding restructuring -FCFM of 31% -Bought back 4% of shares outstanding – $1b. 4QG -Cloud grows 25.5% -OM of 4.5% GAAP and 30.5% PF – reinvesting some of layoff savings into AI and enterprise sales -See a lot of demand opportunity – going to invest in this but while expanding margins. 26G -Rev grows 24% -Cloud grows 26.5% - expect msd contribution from on-prem transition -Data center grows 21.5% -84.5% GM --2% GAAP OM

@Curiousjorge65 Latest quarter includes an acquisition and they guided to exiting the year with flat head count relative to 4Q25. That's before the Armis acquisition so net net will be layoffs.

@TechFundies Lmao look at NOW headcount latest Q, Bill not speaking truths

The number of companies basically saying head count is flat from here on out is pretty shocking. SPOT, NOW, MSFT, META, ... These are big employers... And not exactly like there are a bunch of 50-65 year olds working there who are ready for retirement that would open up seats.

Some highlights from Stripe Sessions. The company now processes >1.5% of global GDP - kind of wild. Ton of AI revenue flows through them as well for payment processing / fraud. Pretty straightforward to gain exposure here - active private market [I picked up a big slug when Harvard was puking stock at 20x FCF in the tariff drawdown last year]. $RVI also has some exposure. Highlights -New company formation on fire - see pic 1 -AI agents interacting with Stripe at exponential rate [think this is people using AI to code integrations for payment processing] - see pic 2 -Comprehensive services – credit card, ACH, stable-coin, fraud-detection, physical terminals, etc. -Big focus on enabling agentic commerce ecosystem – authentication, privacy, etc. -They are processing payments for basically every AI company – Open.AI, Anthropic, etc. -Lots of AI fraud out there basically trying to use tokens for free. 1 out of 6 new account sign-ups are multi-account abuse [wonder if the model companies are deduplicating these KPIs?]. One customer was giving $500 of free AI tokens for trials and only had 4% conversion of trials to new accounts w/ some fraud within that. - see pic 3 -AI went from reading 5% of Stripe documentation for AI implementations to >40%.

@SeattleWA_Mike @SehrLangsam2 Not sure the answer will be either / or. Agents within platforms make sense as do agents outside that do other work. The modern economy is based on specialization so why wouldn't the vendors closest to company departments' workflows be well positioned to enhance productivity?

@SehrLangsam2 @TechFundies What makes you think they will use the SaaS company's agents and not create them with Anthropic or OpenAI?

I swear I'm going to pause posting about AI / SaaS but had to share this A16Z post which I think does a great job of outlining the $WDAY bull case - while respecting their pitch. a16z.com/workdays-last-…

Agreed. SaaS will grow more slowly than in the past. But that doesn't mean growth is going to 0 and that profits can't grow far more quickly with better operating discipline. Multiples will not go back to where they were at prior peaks. But currently valuations can go up by 50-100% and be well below prior averages / peaks.

@TechFundies We need to separate continued revenue growth from stock price. SaaS companies won’t have the same multiples they had in the past. Customers will need to trim their SaaS budgets to move those dollars to AI. SaaS revenue won’t grow at the same rate it did the past 15 years.

@Stormshieldval Anyone could. That's kind of the point.

@TechFundies Pretty sure I could run WDAY margins at double their current rate, and still eek out a bit of growth

Few post-earnings thoughts on $NOW. Then I'm going to take a break from researching / thinking about SaaS after this:) 1) Net new ACV is growing and sales productivity is high (0.6x on trailing 24 month sales expense vs 2025 incremental subscription revenue). This is elite positioning within SaaS where I don't think this can be said of 99% of the sector. 2) Profitability is ramping very quickly. Mgmt guided to their keeping headcount flat yy. Headcount grew 2% qq in Q1 due to MoveWorks and likely grows ~3% qq in Q2 due to Armis. However, it shrinks in the back half. Notably, GAAP opex / head only grew 2% in 2023, 3% in 2024 and 1% in 2025 - pretty respectable. I model 2% yy HC growth in Q4 along with 6% wage inflation. This spits out 22% GAAP operating margins versus 12% in 4Q25. So the setup appears to be this entrenched SaaS operational workhorse, with rational acquisitions that expand TAM and AI enablement with ramping traction, ough to somewhat sustain growth rates (I don't expect an acceleration - just too hard). Meanwhile, the company may emulate the hyperscaler employment strategy of the past five years wherein headcount remains flattish and margins, which are already comically low relative to potential, ramp materially. Stock trades at 26x my 2027 GAAP EPS for a high-retention, recurring-revenue company that ought to grow revenue at 15-20% for a few years with GAAP EPS growing >30%. I also would point out that FCF is material and the company is insulated from the AI capex cycle (technically a beneficiary of it).