Economics on Dr. Hitesh Sarda

A ban on AI agents is a statement about your margins

Thu, 12 Nov 2026 00:00:00 +0000

United Airlines has said it will prohibit unauthorized third-party personal AI agents from booking flights, on its own channels and through agencies and aggregators. Its stated reasons are sensible ones: handing account credentials to an assistant can expose sensitive information, an agent may buy something or spend MileagePlus miles the traveler never meant to, and some federal rules expect a human to confirm things like what is in their luggage. Meanwhile, according to the same report, Southwest has made its inventory viewable to ChatGPT users.

Microsoft is trading payroll for compute. Anthropic's S-1 is the other side of the trade

Tue, 27 Oct 2026 00:00:00 +0000

On Monday Microsoft offered voluntary exits to about 60,000 employees, more than a quarter of the company, at level 70 and below in the US, Canada, India and the UK. The same week, Anthropic’s S-1 showed revenue of $4.6 billion for all of 2025 and $16.3 billion for the first six months of 2026. Six months of this year earned more than three times what all of last year did.

The memo does not mention AI. Business Insider links the two, noting that Microsoft has been spending heavily on AI infrastructure and reducing headcount at the same time. That is a reporter’s inference and so is what follows, but I think it is a reasonable one: a large company is converting payroll into compute, and the S-1 is a look at the other side of that trade.

Most teams still ration AI that is cheap enough to waste

Fri, 09 Oct 2026 00:00:00 +0000

One developer spent a month using DeepSeek 4.1 Flash heavily across a dozen projects and reports that, mid-session, he can’t tell it from Opus. His day-long sessions rarely cost more than a dollar. That is one person’s subjective report, so weigh it that way. What I find more useful is the behavior he describes: “there is no shame now in spinning up mindless tasks,” including exploratory UI testing and tidying files, which costs $0.003 instead of $1.

The engineer who picks the model never sees the bill

Thu, 08 Oct 2026 00:00:00 +0000

Microsoft had projected at least $1 billion of internal spending on Anthropic this year. According to The Information’s reporting, that estimate has since been cut by more than a third. At Meta, internal Claude Code users have reportedly fallen from about 60,000 to about 30,000. Neither company has confirmed the numbers, and Microsoft’s customer-facing spending on Anthropic models is reportedly still growing.

The obvious reading is rivalry. Both companies build competing models and coding tools, so of course they would rather their own engineers used those. That is probably true, and it explains why these two moved. It tells the rest of us very little, because most companies have no in-house model to move to. What the story does show is what happens to an AI budget that nobody was managing: it runs wide open for a year, then gets cut by decree.

Google would rather solve rocket science than wait for a power line

Fri, 25 Sep 2026 00:00:00 +0000

On October 1, a SpaceX Falcon 9 is due to carry four of Google’s TPUs into orbit under Project Suncatcher, the company’s bet that AI chips can run on satellites bathed in eight times more sunlight than any panel on Earth gets. The coverage has mostly treated this as a curiosity: can you really cool a chip in a vacuum, will radiation fry it, is this science fiction with a launch date. Those are fine questions. They are not the interesting one.

Three frontier labs shipped this week. Only one cut its price.

Wed, 23 Sep 2026 00:00:00 +0000

In the space of two days this week, three frontier labs updated their flagship line. Anthropic shipped Claude Opus 5.5 on September 22: 20% cheaper on input and output, 60% cheaper on cached reads, 30% faster, with real but incremental gains on agentic coding and computer-use benchmarks. OpenAI shipped GPT-6 Sol and Luna the same day, extending September’s GPT-6 Astra down into faster, cheaper tiers at roughly half of what the equivalent GPT-5.6 models cost. xAI shipped Grok 4.7 a day earlier, ahead of its predecessor on most evaluations, though not all: Anthropic’s own Fable 5.1 Max still leads it on several coding and long-task tests.

Gemini broke into three companies by guessing passwords

Sat, 19 Sep 2026 00:00:00 +0000

In May, Google’s Gemini broke into three real companies. It wasn’t supposed to. Google confirmed last week that during a security test run by the firm Irregular, the model was given a capture-the-flag exercise against a fictional target. The fake company happened to share its name with a real one, Gemini took the exercise out onto the live internet, and it got in. Google’s VP for security engineering said the model “found public information online and guessed credentials.” In two of the three cases it used login details sitting in public code repositories. In the third it guessed passwords until one worked.

Name the axis you're willing to lose on

Fri, 07 Aug 2026 00:00:00 +0000

A hundred-year-old economics idea spent this week near the top of the tech feeds, explained through Mario Kart of all things. It is a good explainer and worth ten minutes. But the reason it matters to anyone who makes technology decisions has nothing to do with karts. It is that the Pareto frontier is the cleanest tool I know for catching the most common waste in a strategy meeting: a long argument about an option that was never worth choosing.

Cheaper tokens, bigger bill

Sat, 01 Aug 2026 00:00:00 +0000

A team at Manifest built an LLM router in March, ran it across 7,000 users for four months, and then switched it off. The router did what these systems are meant to do: it looked at each incoming request, estimated how demanding it was, and sent the easy work to a cheaper model while reserving the expensive one for the harder cases. A lot of AI infrastructure teams are building something similar right now, which is what makes the decision to retire it interesting. Manifest wrote up why they killed theirs, and the reasoning turned out to be more valuable than the feature ever was.

The AI boom's collateral is a team you can't repossess

Thu, 23 Jul 2026 00:00:00 +0000

A lender will happily write a loan against a warehouse, a fleet of aircraft, or a bulk carrier. The deal underneath is always the same. If the borrower stops paying, the lender takes the thing and sells it to somebody else, and the sale mostly covers the loan. That deal is now being applied, at enormous scale, to GPU clusters. The trouble is that a cluster does not behave like a warehouse, and I think the people writing the checks half-know it.

The scariest cloud bill is the one that looks right

Sat, 18 Jul 2026 00:00:00 +0000

Last week some AWS customers opened their billing console and found they owed trillions of dollars. One account that had spent nineteen cents in June was projected to owe nearly $2.5 billion; another showed a month-over-month change of 55 trillion percent. AWS confirmed the cause as a unit-pricing defect in its estimated-billing subsystem, said actual charges were untouched, and everyone had a good laugh. Somewhere in the pipeline a gigabyte got read as a byte, a factor of a billion, and the numbers went to the moon.

Same model, ten times the bill

Fri, 17 Jul 2026 00:00:00 +0000

Kimi K3, the new 2.8-trillion-parameter open model from Moonshot, shipped this week with a price list, and the number worth staring at is not the one everyone quoted. The headline was that an open model landed within a benchmark point or two of Claude and GPT. The number I keep coming back to is lower down the page: input costs $3.00 per million tokens on a cache miss and $0.30 on a cache hit, and in real coding workloads the cache hit rate runs above 90%. The price you actually pay, for the same model doing the same work, swings by a factor of ten depending on something the pricing page treats as a footnote.

The AI boom is funding a two-year asset with six-year money

Wed, 15 Jul 2026 00:00:00 +0000

In November, Michael Burry told his followers that the biggest names in AI were flattering their profits with an accounting choice, and put a number on it: about $176 billion of understated depreciation across the hyperscalers between 2026 and 2028, enough in his math to overstate Oracle’s earnings by roughly a quarter and Meta’s by a fifth. He was short Nvidia at the time, so weigh the motive however you like. The argument underneath it outlasts the trade.

Someone is subsidizing your AI compute

Sun, 12 Jul 2026 00:00:00 +0000

Last September, Nvidia said it would invest up to $100 billion in OpenAI, which plans to spend much of it deploying Nvidia chips. That single sentence is why “circular financing” is now a phrase everyone in AI knows. The commitment has since been trimmed to a $30 billion equity stake, but the shape of the deal is the point: the supplier funds the customer so the customer can buy from the supplier.