Is your AI spend paying for itself?
You can recite the token bill to the dollar. What the spend returned is the half almost nobody publishes, and you can put a number on it this quarter.
Brian Armstrong published a good cost story this month. Coinbase cut its AI spend nearly in half while token usage kept climbing. The engineering behind it is sound. Smarter model routing, so each task runs on the cheapest model that can actually do it. Aggressive caching, so repeated queries stop paying for redundant outputs. A shift of routine work onto cheaper open-weight models, where the frontier adds nothing.
He goes further with a prediction. Within twelve to eighteen months, he argues, roughly 80% of workloads will run on models that cost about 99% less than today's frontier. The flagship models get reserved for the hard problems: scientific breakthroughs, complex agent orchestration. In his telling, the binding constraint shifts from better models to energy and compute. The shift now has a name circulating widely in the press, the end of the tokenmaxxing era, though Armstrong himself did not use the phrase.
I have no quarrel with the engineering. Coinbase was reportedly generating about 40% of its daily code with AI as of late 2025, with a stated goal to pass half, so for them this is shipping practice.
It does. Cheap inference is the most reliable trend in computing right now. Since ChatGPT shipped in November 2022, the cost of reaching a fixed level of capability has fallen on the order of 10x a year. The 2025 to 2026 price war alone knocked mainstream API prices down 60% to 80% in a single year. So 99% cheaper is a 100x drop, and the curve keeps clearing that bar in well under two years. Armstrong's twelve-to-eighteen-month window is a forecast, and an aggressive one, but it sits inside the trend. The claim that 80% of workloads commoditize is just the rule that yesterday's frontier becomes today's default. We watched it happen to GPT-3.5, then to GPT-4. The frontier itself can still get more expensive: OpenAI roughly doubled its top-line price when GPT-5.5 shipped in April 2026. That fits the rule. The curve collapses the cost of a fixed capability, while the new frontier, the genuinely hard 20%, stays premium.
Most of that decline comes from software and markets. Algorithmic efficiency improves around 3x a year. Mixture-of-experts designs and better inference systems compound it, and an oversupply price war is doing much of the rest. Raw hardware cost per unit of compute, by contrast, improves only around 1.3x a year. The fall is also uneven. Routine work deflates fastest, and that is the easy 80%. The genuine frontier stays expensive, which is why the flagships get reserved for it, and why a new flagship can still cost more than the one it replaced. You can watch the split in a live market. OpenRouter routes across hundreds of models and publishes token volume by model, and outside analysts who layer in price find a familiar shape: a premium lab pulling a large share of revenue on a small share of the tokens while cheap open-weight models carry the bulk. That is the 80-20 split showing up as market structure.
This is uncomfortable news for anyone bragging about a cost cut. A deflation this predictable and this exogenous arrives for every competitor on the same schedule, so it gives nobody a moat. If your token bill falls 90% next year, so does theirs. Riding the curve well, with good routing and caching, can buy a temporary edge. The deflation underneath belongs to everyone. What you do with the cheap tokens, the return, is the only part you durably control.
So cost is one side of the equation: the denominator. Almost every public story about AI spend right now leads with it. That choice shapes companies, because what leadership publishes is what the org builds dashboards around.
02 · The numeratorThe return number your AI budget never shows you
The return side mostly stays in the drawer. Only 27% of CEOs say AI has met or exceeded their ROI expectations, down from 38% the year before. That figure comes from the Oliver Wyman Forum's CEO Agenda 2026, a survey of 415 chief executives, the people who sign the checks. And 53%, an outright majority, say it is simply too early to tell. The group that can point to a clear, met return is small and shrinking. The largest group cannot see the return at all yet.
A CEO whose AI "has not met expectations" may still see the return clearly. Some of those CEOs measure cleanly and dislike the figure. Some set the bar too high at the start. The 53% who say it is too early to tell are describing something specific: nine months and millions of tokens in, they still cannot put a number on what the spend returned. One trap catches all three groups. If you cannot see return by workflow, you cannot tell "AI genuinely underdelivered here" apart from "I just cut the one workflow that was actually working." Measurement settles that question whichever camp you are in, which is why it comes first.
Cost is being optimized hard, in public, with case studies and threads and crisp before-and-after charts. Return is barely measured at all, and return is the half of the equation that tells you whether the spend was worth making.
The appetite is there. In one widely cited industry survey, 85% of organizations want to be agentic within three years and 76% say their current operating model cannot support it (MIT Technology Review Insights, May 2026, sponsored research in partnership with Ema, with the 76% figure drawn from Celonis). Treat those figures as directional, given who funded them. The direction rings true, though, and it points at what an operating model that "cannot support" agentic AI is missing. You cannot redesign around the workflows worth running if you have never measured which ones return. Much of the readiness gap comes down to measurement.
The same token spend can produce a return of negative one hundred percent or positive ten thousand percent, depending entirely on what you do with it. Cost tells you nothing about which one you have. A team that cut its AI bill in half may have trimmed the workflow returning thirty times its cost while lovingly protecting the one returning nothing. The bill went down, the loss never showed up on any report, and nobody noticed, because nobody was watching the return.
The marketing move that makes your AI spend pay for itself
Marketers solved a version of this problem a long time ago, and they gave it a name.
A self-liquidating offer is a front-end offer whose revenue covers, or liquidates, the cost of acquiring the customer. The acquisition pays for itself. The instant that happens, good marketers scale the ad spend as hard as the offer will hold. Why would you cap something that returns more than a dollar for every dollar in, plus a customer on the back end? From that point, cost is a throttle you open.
Transplant the idea straight across. Token spend is self-liquidating when the measurable value it produces covers its fully-loaded cost. Once a workflow clears that bar, you feed it more spend, in steps. A workflow that liquidates itself at five thousand prospects a month can behave differently at fifty thousand, as rework creeps up and the list saturates, so each order of magnitude earns its own quick test before the next. With the numerator known, the frontier-versus-cheap-model debate becomes routine tuning of the denominator.
The analogy strains in one place. Attribution is harder here than in direct response. A coupon code ties a sale to an ad with near-certainty. A workflow's contribution to revenue travels through more hands: the rep, the offer, the brand, the timing. That difficulty shapes the value ladder below, which starts at the rung you can defend cold and climbs only when the evidence earns it.
So the operating question for any founder becomes: is my token spend paying for itself, and how would I know?
Put a return number on every AI workflow you run
Start with one metric and one ratio.
Token ROI (tROI) = (value produced minus fully-loaded token cost) divided by fully-loaded token cost. Self-Liquidation Ratio (SLR) = value produced divided by fully-loaded token cost. That is the friendly form, the one you can say out loud in a meeting.
- SLR of 1.0 or higher: the spend is self-liquidating. It pays for itself. Open the throttle and scale.
- SLR below 1.0: not yet. Fix the value side, or kill it.
Everything turns on "fully-loaded." It means API and token cost, plus caching and infrastructure, plus the human review and rework time the output actually demands. A 99%-cheaper model whose work a person has to redo can end up costing more than the model it replaced. The quality drop that creates rework is a real cost, and it belongs in the denominator where it can be seen.
None of this needs new vocabulary. The Self-Liquidation Ratio is just the general form. Translate it into the cost metric each team already trusts. MarketingOps lives it as cost of acquisition: you spend to win a customer only when that customer is worth more than the spend, which is the original self-liquidating logic. SalesOps has cost per opportunity. Support has cost per resolved ticket. Engineering has cost per shipped feature. Tokens are just another input cost flowing into each of those lines. Call the ratio whatever your team says out loud. The discipline holds everywhere: every dollar of added cost, token or otherwise, has to produce a valid, measurable business return, or it gets cut.
Marketers learned one more lesson the hard way: a spend number means little without a return number sitting next to it. The bare token bill is a rear-view mirror. It reports what you already paid. The Self-Liquidation Ratio works like a windshield. It estimates what the spend will earn back, the same way a marketer weighs the cost of winning a customer against the gross profit that customer throws off over a lifetime. Publish only the rear-view number and you are steering a fast car by the road behind it. The windshield depends on one thing: a number for what each workflow returns.
The value attribution ladder
People skip measuring return because "value produced" feels impossible to pin down. A ladder makes it tractable. You attribute at the highest rung you can defend, you discount for confidence, and you never count a rung you cannot evidence.
| Rung | What it measures | How to value it |
|---|---|---|
| 1 Displaced cost | The spend replaced a known cost: an FTE-hour, an outsourced deliverable, a SaaS seat | The cost you no longer pay. Easiest, most defensible |
| 2 Throughput | Same headcount, more output or a faster cycle | Incremental margin from extra output, or the dollar value of compressed time-to-cash |
| 3 Revenue lift | The spend produced or influenced revenue directly | Attributed revenue times margin, discounted by attribution confidence |
| 4 Optionality | Capabilities, moats, proprietary data, learning | Track qualitatively with a confidence flag. Never launder it into a hard number |
Rung 1 is where you start, and it is more than enough to begin. A workflow that produces twenty-five thousand custom drawings in three minutes, replacing roughly five full-time people, has a value you can defend to a CFO without hand-waving. A support assistant that deflects forty tickets a month, against a blended loaded cost of a few hundred dollars, has a rung-1 number you can post the first week it runs. The two common mistakes mirror each other. One is claiming rung 3 or 4 with no evidence. The other is refusing to measure anything because rung 3 is hard, while rung 1 sat in plain sight the whole time.
The middle rungs are just as defensible when you anchor them to a number you already keep. A contract-review workflow that clears agreements 40% faster may never touch an attributable sale. Its compressed cycle time still has a dollar value: deals signing sooner, cash landing sooner, a backlog clearing without another hire. A CFO will accept that rung-2 figure.
The four numbers that turn the ratio into a system
The ratio tells you where a workflow stands. Four supporting numbers tell you what to do about it. Cost per outcome, what you pay for each accepted unit of value. Quality-adjusted cost, which is token cost plus rework cost, and which exposes a cheap model whose rework makes it expensive. Payback velocity, how fast an initiative crosses SLR 1.0. And the yield curve by workload, the one that changes how you run the company.
Token ROI: net return over fully-loaded cost
Self-Liquidation Ratio: value over cost
Cost per accepted unit of value
Token cost plus rework cost
How fast a workflow crosses SLR 1.0
tROI and SLR by workload
Cost per outcome is the true denominator, quality-adjusted cost is the lie detector, payback velocity is the clock, and the yield curve is the map. Together they turn one ratio into an operating instrument.
05 · The portfolioWhich of your workflows to scale, fix, optimize, or kill
Once you have a yield curve, you can manage AI spend the way an investor manages a portfolio. Some workloads are thirty times self-liquidating. Some are negative one hundred percent. A single company-wide "AI ROI" number hides every bit of that, and so do the headline figures most companies report instead: total tokens consumed, or the bare token bill. Both are aggregates, blind to which workflow earned its keep and which one quietly bled. Plot every material workload on two axes, return and cost, and each one's next move becomes obvious.
| Low cost | High cost | |
|---|---|---|
| High return | Scale Self-liquidating. Cost is a throttle. Open it. |
Optimize Proven return. Now apply routing and caching, here. |
| Low return | Fix or kill The vanity trap. Busy, cheap, pointless. |
Kill Bleeding. Kill it fast. |
The high-return boxes are where conventional wisdom breaks. Reflexive cost-cutting there is a mistake. Trimming the spend on a workflow that returns thirty times its cost starves your single best asset so the report looks disciplined, and the applause for the smaller bill hides the loss. It is the founder's oldest trap, the strength that built the business turning into the constraint that holds it back.
06 · The synthesisWhere the cost playbook belongs
Look at the matrix again. The routing-and-caching playbook belongs in the high-return, high-cost box, and it earns its keep there. The return is proven, so that is the place to optimize the denominator. You already know the numerator is real.
Pair every cost cut with a return number.
Armstrong's playbook is sound. The trouble starts when it stands in for return measurement. Routing and caching can run in parallel with that measurement, and often should. Nobody has to prove every return before touching cost. But optimizing the cost of unmeasured spend is optimizing in the dark. Routing and caching make a returning workflow return more efficiently. They say nothing about whether it returns at all. Run them as your only discipline and you will dutifully, efficiently, scale your way deeper into the group that cannot see its return.
One fair pushback: measurement costs time too, and a founder's bandwidth is finite. So tier it. Point your instrumentation at the highest-cost workflows first, the ones where a wrong call costs the most, and let cheap heuristics ride on the small, uncertain spend until it earns a closer look. You can stop cutting blind without measuring everything at once.
07 · The litmusA three-check test for token return
For each material workload, can you state its SLR? If you cannot, you are measuring tokens and guessing at return. Three flags say you are not there yet.
- Leaderboard gap. You can quote your token spend or your cost-cut percentage. Your token return has no number.
- Attribution gap. None of your top three workloads has a value-attribution rung assigned to it.
- Aggregate blindness. You report one company-wide number, total tokens used or the bare token bill or a single "AI ROI" figure, with no per-workload yield curve underneath it.
How to instrument it
It is less work than it sounds, because half the rig already exists.
- The cost meter is already wired. Model-routing infrastructure of the kind behind Armstrong's cost cut emits per-call metadata. The value meter is usually missing. Tag spend at the workflow level, where return can be matched to it.
- Pick the highest defensible attribution rung per workflow, and write down the confidence.
- Emit an accepted-outcome event from every material AI workflow. Cost divided by accepted outcomes gives you cost per outcome, for free.
- Run a monthly yield-curve review. Scale the top quartile. Fix or kill the bottom.
- Report cost per outcome and SLR to the board and the team, in place of total spend and token volume.
A worked example, the winner. Take an AI-assisted outbound sales workflow, measured over one month.
| Cost, fully-loaded (one month) | Figure |
|---|---|
| Tokens and API: about 5,000 prospects run multi-step (research, draft, two follow-ups), roughly 100 million tokens at a blended $6 per million | $600 |
| Orchestration, enrichment data, caching | $200 |
| Human review and approval, about 13 hours of a rep's time | $700 |
| Fully-loaded cost | $1,500 |
| Value, attributed gross profit (one month) | Figure |
|---|---|
| 5,000 personalized emails, 0.8% book a meeting | 40 meetings |
| 25% become qualified opportunities | 10 opps |
| 20% of opportunities close | 2 deals |
| Gross profit per deal ($15,000 average deal at 60% margin) | $9,000 |
| Gross profit produced | $18,000 |
| Attributed to the workflow, a conservative 50%, since the rep and the offer also worked | $9,000 |
SLR = $9,000 / $1,500 = 6.0. tROI = 500%. Cost per closed deal = $750, against $9,000 of gross profit on each one. The bare token bill was $600, and that is the number most teams would publish. On its own it says nothing about whether the workflow worked. The fully-loaded ratio says scale it, and the next dollar belongs in more prospects and a better research model.
A word on that 50%. It is a placeholder for a test you have not run yet. To earn a real attribution figure, run the same lead set through an AI-assisted rep and an unassisted one and watch the gap. Until then, 50% is a conservative anchor, and the rule is to round down. The real number will catch up with you either way.
Now run it on an AI-assisted competitive-research workflow. A strategist spends three hours a week prompting, reading, and verifying outputs that mostly confirm what the team already suspected. Token cost is modest, call it $120 a month. Loaded with the strategist's time, the cost is closer to $900. Rung-1 displacement: zero, because nothing was being outsourced before. Rung-2 throughput: marginal, because no decision moved faster. Rung-3 revenue: none anyone can trace. SLR lands well under 0.5. This is the bottom-left box. Cheaper tokens cannot rescue it, because the value side is empty. Redesign it toward a rung-1 displacement that actually exists, or kill it. The cheap-model reflex would have quietly kept it alive.
The sharpest version of cost per outcome is one almost nobody tracks: cost per accepted change, meaning the cost of the work that actually cleared for production. The threshold is intuitive once you name it. If a loop hands you ten results and you throw six away, you are doing the review work the tool was supposed to save, and below roughly a 50% accept rate it can cost more than it gives back. The idea started with coding agents, and it generalizes to every loop a business runs: drafts a rep sends versus drafts they rewrite, support replies that ship versus replies redone. Acceptance rate is where token cost meets human rework. It counts only the work that survived the bar, which makes it the truest read on productivity. Fold it into the denominator. A falling accept rate is the early warning that a cheaper model is quietly costing you more. (The framing is Anatoli Kopadze's. The rough 50% line is my own working threshold, an estimate from practice.)
If your CFO cannot calculate your cost of inference per completed workflow, the architecture is not yet operationalized. Tokens are cost of goods sold. Treat them like it.
What to publish this quarter
Cost discipline without return measurement is thrift: useful, and blind. The companies that win the next eighteen months will know, per workflow, exactly what each token returned. They will feed the spend that pays for itself and kill the spend that does not. Cost optimization makes a company efficient. Return measurement tells it where efficiency is worth buying. Pair them, return first, and the cost work finally has a target worth aiming at.
The hard edge cuts both ways. Some workflows will never look token-return optimal, and next to the ones that do, they will read as expensive. Be careful what you call a loser. The cost curve is still falling, so a workflow that is underwater today can surface in a year without changing at all, the denominator dropping out from under it. Some value is real and refuses to become a number, which is exactly what rung 4 is for, and a spreadsheet's limits are no reason to kill it. And some buyers have so much money that price never enters the decision, so for them the return question was never about cost.
Underneath all of it sits an open question. As inference trends toward free, does the self-liquidation test stay binding, or does almost everything eventually pay for itself, leaving judgment, taste, and trust as the only things left to compete on? I do not know. Nobody does yet. That is the part worth watching.
Yield is the number to publish.
Publish your token-return KPIs: cost per outcome, SLR by workload, your yield curve. For the majority of leaders who still cannot see their return, that post would help far more than another token-spend brag. Pick your top three AI workloads this week. For each one, answer a single question: is this spend paying for itself, and how do I know?
Sources
- Brian Armstrong, post on X, June 2026: x.com/brian_armstrong
- Tekedia: "Coinbase CEO Brian Armstrong Urges Shift to Cheaper AI Models, Signaling End of the 'Tokenmaxxing' Era" (the "end of the tokenmaxxing era" phrasing is the outlet's own framing; Armstrong did not use it)
- The Block: "Brian Armstrong says about 40% of Coinbase's daily code is AI-generated" (figure as of late 2025, with a stated goal to pass 50%)
- Oliver Wyman Forum, CEO Agenda 2026 (survey of 415 CEOs: 27% say AI met or exceeded ROI expectations, down from 38% the prior year, with a majority saying it is too early to assess)
- MIT Technology Review Insights, May 2026, "Rethinking organizational design in the age of agentic AI" (sponsored research, in partnership with Ema; the 85% agentic-intent and 76% operating-model figures, the latter drawn from Celonis)
- Stanford HAI, AI Index Report 2025 (GPT-3.5-level inference cost fell about 280x, November 2022 to October 2024)
- Andreessen Horowitz (Guido Appenzeller): "Welcome to LLMflation", November 2024 (about 10x per year, near 1000x in three years)
- OpenAI API pricing (2026): GPT-5.5 flagship at $5 / $30 per million input/output tokens, roughly doubling the top line at the April 23, 2026 release; mainstream API prices fell roughly 60% to 80% across 2025 to 2026 (the AI price war)
- Epoch AI: "LLM inference prices have fallen rapidly but unequally across tasks" (per-tier range roughly 9x to 900x per year, median near 50x; GPT-4-level output fell roughly 40x cumulatively off a March 2023 launch price near $60 per million; algorithmic efficiency about 3x per year versus hardware about 1.3x per year)
- OpenRouter, market-level usage across hundreds of models (token volume published officially at openrouter.ai/data; per-model revenue shares come from third-party analysts layering price onto that volume, and OpenRouter does not publish them)
- Anatoli Kopadze (X, 2026): the cost-per-accepted-change framing and the rough 50% accept-rate threshold, used here as a springboard; the threshold is the author's synthesis. x.com/AnatoliKopadze
- Jeremy Kahn, Fortune (May 2026) and Azeem Azhar with Nathan Warren, Exponential View (May 2026), on the "tokenmaxxing is over" discourse
- Founder OS, building-an-exo and analyzing-ai-roi skills (analyzing-ai-roi absorbed the earlier measuring-token-roi skill): the tokens-as-COGS framing, cost of inference per completed workflow
Notes like this one also go out through factually, my newsletter. Subscribe at news.kentlangley.com, or point your reader at the RSS feed.