Uber's $1,500 Cap: Repricing Your AI SaaS Before Buyers Do It for You

Derek SimmonsOkay, real talk for a second. Uber blew its entire 2026 AI coding budget in four months.

Elena ReyesFour months?

Derek SimmonsBloomberg confirmed it.

Elena ReyesWow.

Derek SimmonsAccording to Simon Willison.net, they've now capped every engineer to $1,500 per tool per month. 95% of their 5,000 engineers were actively burning tokens. And then GitHub Copilot flips to usage-based billing on June 1st. And according to Ars Technica, some devs... Devs burn through their whole monthly credit allotment in a single day. A single day. One day. So what does this mean for the person listening right now? If you're selling SaaS into enterprise buyers, finance teams are now line item managing AI spend. That is a completely different procurement conversation than it was 12 months ago. And that is exactly where today's guest walked straight into this mess. We're going to reconstruct their ARR, their buyer profile. While the specific deal that got killed by an AI budget objection and how it compared against the 2026 mid-market close rate benchmark of 20-28%. And then it gets good. Oh, it gets good. We dig into the pricing page changes they made after this whole AI budget categorization problem surfaced, before and after conversion data, a shift toward a subscription floor plus usage upside structure, and whether outcome-based pricing is is actually closing deals at the $200K to $5M ARR stage, or just appearing in decks, or just appearing in decks, and we close with what broke operationally when the new pricing motion actually had to scale, plus a concrete action founders should take before their Q3 renewal calls arrive. There is a lot to get through. Let's get into it. All right. First up, the data drop. Okay, so get this. Uber burned through its entire 2026 AI coding budget in four months. Four months! Then capped every engineer at $1,500 per tool per month. Per tool? Per tool. So if you're running Cursor and Claude Code, that's $3,600 a month per engineer, max. Bloomberg reported it, Simon Willison unpicked it apart. And before the cap, what were individual engineers actually burning? According to Yahoo Finance's reporting, some were hitting $500 to $2,000 in token consumption, and the Uber CTO reportedly torched 1,200 bucks in tokens during a two-hour internal demo. $1,200 in a demo? For a demo. So Finance noticed. Right, right. But here's the thing that caught me, Derek. GitHub Copilot flipped to usage-based billing on June 1st. Ars Technica covered the immediate fallout.

Speaker 3And?

Derek SimmonsOne Pro Plus subscriber $39 plan burned through 8% of their entire monthly credit quota in two hours. Normal workflows.

Speaker 3Wait, wait. 8% in two hours?

Derek SimmonsTheir own words, the 7,000 unit quota gone in under two days at that rate. So they went from a predictable subscription to watching a meter spin like a taxi in Midtown. down. And people are furious, the register covered developers threatening to just leave Copilot entirely.

Speaker 3And here's why this matters to the person listening right now, though. These are not edge cases. Uber has 5,000 engineers, 95% of them using AI tools every month. That is a finance line item that got out of control fast. So enterprise procurement teams are now actively managing AI spend, not passively.

Derek SimmonsThey have dashboards, they have caps, they have approval workflows to exceed those caps. And every SaaS founder selling into that buyer in 2026, your pricing page was probably built before any of this existed. Seriously, the 2026 SaaS benchmarks report from SaaSrise pulls data across 2,500 companies. The AI budget conversation is showing up as a procurement variable that most pricing models were not designed for.

Speaker 3Four. Twist twist. The objection in your next enterprise deal isn't your price is too high. It's your product lives in the AI tool budget bucket and that bucket has a hard cap.

Speaker 4So the question is, when did you first hear buyers actually say those words? And what did the deal look like when they did?

Derek SimmonsSo here's where the rubber meets the road. Walk us back to what the business looked like before this started. ARR, team, and the kind of buyer you were closing.

Speaker 4Yeah, and I want a real number, not mid-market. What were you actually at?

Derek SimmonsRight, because the shape of the problem changes a lot depending on whether you're at 400K or at 2M.

Speaker 4Completely different conversation.

Derek SimmonsSo give us the snapshot.

Speaker 4And specifically, what did your buyer profile look like? Like who is signing these deals?

Speaker 3Okay, so let me push on the moment, not the quarter, the moment. When did you first hear this objection on an actual call? What did the buyer say?

Speaker 4Because I want to reconstruct the exact language, not we have budget concerns. That's not what they said, right? No, no, no. Budget concerns is the old objection. This one sounds different. The new version sounds more like we have a line item for AI tools and your product lives. Lives there.

Speaker 3And that line item has a number on it.

Speaker 4Which is the whole problem. So what number did they give you? Do they actually say a figure? And did it stall a deal or kill it outright? Those are two very different case files. Exactly. Stalled means there's still a path. Killed means the budget bucket closed before you even got to a second call.

Speaker 3I've seen this movie before. The deal looks alive for six weeks, and then you find out it was dead at week two, they just hadn't told you yet. The zombie deal pipeline looks full, close rate collapses, which actually brings me to the number I need. What was your close rate the quarter before this objection started showing up versus after? Give me a percentage, not a feeling.

Speaker 4Okay, but let me stress test that for a second. According to 2026 benchmarks, mid-market SaaS close rates run around 20 to 28%, median sitting near 24. So if you were above that, That before and dropped below it after, that's a clean signal.

Speaker 3That's the number. What's your before and after?

Speaker 4Because if you went from, say, twenty six percent down to fourteen percent,

Speaker 3Wow.

Speaker 4that's not noise. That's a structural objection showing up in your pipeline.

Derek SimmonsAnd the SaaS for as 2026 benchmarks data pulled across 2,500 companies would put that kind of drop squarely in something changed in your sales motion territory. Not a bad quarter.

Elena ReyesMm-hmm. Though to be fair, some founders will say their close rate was already soft before, so we need the specific deal. Walk us through one you lost. Not the pattern, the deal.

Derek SimmonsRight. What was the ACV, how hard did it get, and exactly where it stopped.

Elena ReyesAnd here's what I... But I actually want to know. Did the buyer tell you which AI tools were already in that bucket competing for the same spend?

Derek SimmonsBecause that changes your counter entirely. If you're sitting next to Copilot or Claude in someone's AI tool line item, that's a very specific problem.

Elena ReyesYeah, yeah, it tells you what category the buyer has mentally filed you in, which is often wrong.

Derek SimmonsThat's the real damage. The categorization problem is harder to fix than the price objection.

Elena ReyesAnd that's actually where the pricing page conversation starts. Because if a buyer is miscategorizing you, the page is often the first place they form that mental model.

Derek SimmonsWhich is exactly what we're going to pull apart next.

Elena ReyesSo, we left off with the pricing page as the root cause. Walk me through what it actually looked like before you changed anything.

Derek SimmonsAnd I want the before number first. Conversion rate, close rate, whatever you were tracking. The data before you touched it.

Elena ReyesRight, because the rationale is easy. Everyone has a rationale. What was the number?

Derek SimmonsWe were struggling is not a number.

Elena ReyesNo, it is not.

Derek SimmonsSo what did the page actually say about AI? Was it a separate line item, a feature bullet? How was it presented?

Elena ReyesBecause that's the categorization problem we flagged. If your pricing page uses the word AI anywhere near a usage or credit metric, you just told every enterprise procurement team exactly which budget bucket to drop you into.

Derek SimmonsThe hard-capped one.

Elena ReyesExactly. And once you're in that bucket, you're competing against Uber's $1,500 per tool cap. cap not against your actual value.

Derek SimmonsOkay, so what changed? Walk me through the specific page at it.

Elena ReyesAnd the 2026 pattern that's actually working, according to OpenView's benchmarks, is a subscription floor with usage upside, base fee that maps to a familiar budget line, then metered AI features on top. Companies running that hybrid structure are reporting NRR roughly six to ten points higher than pure subscription peers.

Derek SimmonsWait, so the base fee isn't just a revenue thing, it's a budget categorization thing.

Elena ReyesThat's the whole move: you get out of the AI tool budget and back into software budget. Same product, different line item on their spreadsheet.

Derek SimmonsHuh, so did that actually shift the close rate?

Elena ReyesThat's the question. What was close rate before, what was it after, and how many deals went through before you had enough data to call it a trend and not luck? Luck.

Derek SimmonsBecause three deals is not a trend. I've seen this before, where a founder changes two things at once and then credits the pricing page.

Elena ReyesSure.

Derek SimmonsYeah, yeah. So let's talk about outcome-based pricing for a second, because I know it came up in your pitch materials. Gartner projects 40% of enterprise SaaS contracts will include outcome-based components by end of 2026, which sounds huge.

Elena ReyesRight, but here's the number nobody puts in the deck. According to NxCodes February 2026 pricing data, only 9% of companies have actually fully implemented it.

Speaker 3So the gap between we're exploring outcome pricing and we have an outcome pricing model is enormous.

Elena ReyesEnormous. Yes, and at 200K to 5M ARR, I want to know, did outcome pricing actually close deals, or did it just sound great in the intro slide?

Speaker 3Real talk, because I've watched founders pitch outcome pricing who cannot tell me how they'd measure the outcome, let alone enforce it in a contract.

Elena ReyesHmm, attribution is the whole problem. If your AI helps close a deal but a human rep was on every call, what did you actually deliver?

Speaker 3And procurement is not signing a contract where outcome is fuzzy.

Elena ReyesNo. So what you probably did, if we're being honest, is land on a hybrid. Hybrid. Base subscription, usage layer on top. Maybe outcome language in the deck to get the meeting, but a hybrid structure to close it. Which brings up the next thing, because the pricing page is only half the fix. What does your rep actually say on the call when the buyer still invokes the AI budget cap even after seeing the new page? That's where the architecture and the live conversation have to match. And that's exactly where we're going next. next

Speaker 3So, the pricing page gets buyers past the first filter. But here's the thing. A rep still picks up the phone and the buyer says, we've already spent our AI tool budget. What do the first 30 seconds sound like?

Elena ReyesRight. And this is where most reps just fold or pivot to a discount. Neither of those works.

Speaker 3Neither works. So walk me through the actual words.

Elena ReyesScript that moved the needle was basically this. Before we go there, I want to ask one question. When you say AI tool budget, are you referring to your IT operations bucket or your software productivity line? That's it, first 30 seconds.

Speaker 3Hold on, that's not a pricing conversation. That's a categorization audit right there on the call.

Elena ReyesExactly, because 90% of CIOs told Zylo that cost forecasting is their top challenge in AI deployment. The buyer doesn't want to say no to your product, they want a reason to move it to a different line item.

Speaker 3item so the rep isn't selling they're handing the buyer a shovel to dig themselves out of the wrong budget bucket that's a much better way to put it okay but let me stress test that what if the buyer says no it's all one pot and that pot is locked fair

Elena ReyesThen you've learned something valuable in 30 seconds instead of 30 days.

Speaker 3but what actually moves close rates back because that script addresses the opening objection what did the dashboard look like the months Things started recovering.

Elena ReyesSo the first metric that moved was meeting to proposal conversion, not close rate. Proposals.

Speaker 3Seriously? Not close rate first?

Elena ReyesNope, because the old objection was killing deals before a proposal even went out. Once the categorization question stopped the bleed, proposals went out. Close rate came back maybe six weeks later.

Speaker 3And what did the team think was working that turned out to be wrong?

Elena ReyesOh, this is.

Derek SimmonsThis is good. They credited the pricing page redesign

Elena ReyesRight.

Derek Simmonslike fully convinced the new page was doing the work.

Elena ReyesPlot twist, it was the call script.

Derek SimmonsIt was the call script. The page got them into the meeting. The 30-second question kept them in the deal.

Elena ReyesYeah. Here's the tension I want to name. According to SoftwareSeni, 43% of enterprise buyers say outcome-based or risk-share pricing matters to them. But buyers also keep asking for flat fees because

Derek SimmonsBecause those map to existing budgets." That contradiction resolved itself on the call when the reps stopped defending the pricing model and started asking which budget line fit. You let procurement solve their own problem. Two things can be true at once: buyers want outcome based in theory; they need predictable in practice. The CALL script bridges that gap without forcing a negotiation. And SaaSrise's twenty twenty six benchmarks back that up. Companies that landed on hybrid structures with a subscription floor saw a Saw NRR recover faster than pure usage place. Which number moved last? Expansion ARR, that was the last to recover. New logo close rate came back first, then NRR, then expansion; sequential, not simultaneous. So the motion worked, and that's where our next question gets uncomfortable, because when a motion works something else breaks. Billing, forecasting, CS handoffs, something always breaks. Yeah, and that's a very different kind of problem than losing deals.

Elena ReyesSo the motion is working, close rates recovered, expansion ARR is moving, what broke downstream?

Derek SimmonsBilling infrastructure, month seven. The subscription floor plus usage upside structure they'd build for the Objection conversation created a forecasting mess their finance team wasn't set up for. Flat fee hits one line, usage variance hits another. Nobody reconciled them until a board meeting.

Elena ReyesClassic. You solve the sales problem and create an accounting problem.

Derek Simmonsproblem. Two for one.

Elena ReyesOkay, but here's the pressure question for founders listening. You fixed pricing. You have renewals coming. Here's the thing that should keep you up at night. Per token prices have fallen roughly 98% from 2022 levels according to multiple industry analyses. Enterprise AI bills are still up an estimated 320%.

Derek SimmonsWow.

Elena ReyesYour buyer's CFO knows this math now.

Derek SimmonsSo buyers are walking into renewal thinking tokens got cheaper. why is my SaaS bill the same?

Elena ReyesExactly. And if you don't have a prepared answer, that's not a pricing conversation, that's a cancellation conversation.

Derek SimmonsHave you actually built a renewal script for that scenario, or are you hoping it doesn't come up?

Elena ReyesHope is not a renewal strategy.

Derek SimmonsOkay, so concrete actions. Someone's listening right now, their pricing page still says AI-powered somewhere, renewals land in Q3, what do they do this week?

Elena ReyesOne thing: pull up your three biggest renewal accounts.

Derek SimmonsMm-hmm.

Elena ReyesLook at how your product is categorized in their spend management tool—Ramp, Brex, whatever they're using. If you're in the AI tools bucket with a hard cap, you're negotiating against Uber's $1,500 limit before the call even starts.

Derek SimmonsSo you need to know which bucket you're in before they tell you on a renewal call.

Elena ReyesBecause by then, procurement has already made the decision. You're just there to hear it.

Derek SimmonsNot great odds.

Elena ReyesRight, that's a wrap on this one, and honestly, if you're a SaaS founder selling into enterprise right now, I hope that objection reframe stuck.

Derek SimmonsThe categorization thing, that's the one I keep coming back to. Your product lands in the AI tools bucket, and according to Simon Willison's coverage of the Uber story, that bucket now has a hard cap.

Elena ReyesFifteen hundred dollars per tool per month,

Speaker 3Mm-hmm.

Elena Reyesand GitHub Copilot's June one billing switch showed buyers are already feeling that burn firsthand.

Derek SimmonsSo the audit question we left listeners with: Check where your product lives in your buyer's spend management tool before Q3 renewals hit.

Speaker 3Yeah.

Derek SimmonsThat's the homework.

Elena ReyesNot glamorous, completely necessary.

Derek SimmonsNever is. If this episode saved you from a bad renewal conversation, share it with one founder who needs it. Is it? Subscribe on YouTube or your podcast app, drop a review so we keep getting guests to share the real numbers. We'll see you next time on ARR Autopsy. Thanks for listening.