Public Markets & IPO

Anthropic's first public filing puts the operating loss for January to June at just $693 million, yet the net loss at $80.7 billion. The gap is accounting for convertible notes and derivatives, not cash burned on GPUs.

Commenters spent their time on a different number: how much of that revenue is one product, and whether anyone can buy in at a trillion-plus valuation and still get Amazon-sized returns.

The author's favorite explanation is not generosity. Models are now swappable through routers like OpenRouter, and 29 hosts already serve one open DeepSeek model.

The thread's most practical answer came from a team that left for cheaper models over refusals.

Slok's chart set also checks total factor productivity from a San Francisco Fed series and finds no AI lift. His read: mainstream companies now treat token spend as a margin cost, not a growth engine.

The thread's counter was a mill with a water wheel: you can't drop an electric motor into it, you rebuild the building.

Commenters questioned both the premise that one company controls AI’s trajectory and the likelihood that the United States would permit a takeover with strategic consequences.

The more concrete alternative was not a purchase: build compute capacity with Norway’s energy resources and compete for talent directly.

Apple says alternative distribution still requires notarization, while qualifying web-distributed apps move from a per-install charge to the Core Technology Commission.

Developers focused on the awkward edge case: a customer downloads an app directly, pays for a digital service, and Apple takes a share despite not processing the sale.

A yuan clearing bank gives institutions a dedicated route for settling renminbi transactions. Deutsche Bank's new role makes that route more accessible from Europe.

HN turned the announcement into a larger argument about reserve currencies: gradual trade plumbing may matter more than dramatic declarations.

The story is about how hyperscalers finance enormous capital spending, and whether lenders will keep absorbing the paper. That is a financing question before it is a technology verdict.

The discussion’s clearest reversal: heavy borrowing and a debt crisis are not synonyms, especially when the article itself describes investor appetite alongside the warnings.

Apple Upgrade leases iPhones for 12 or 24 months and now extends the same model to Macs, iPads, and Watches.

The change is not just a new monthly-payment label: keeping the device requires a separate purchase option at the end.

The page that just went live promises ads "clearly labeled" and "separate from answers," rolling out first to only seven countries, about 8% of the world's population, and skipping every paying subscriber.

The asterisk sits in OpenAI's own targeting rules: sensitive chats stay ad-free, but the memory they leave behind doesn't. It can still steer which ad shows up in an unrelated chat the next morning.

A Nikkei investigation puts five US tech giants' off-balance-sheet AI debt at $1.65 trillion, an eightfold jump in four years that now tops the $1.35 trillion they report outright, mostly through data-center leases and GPU deals that never touch a debt line.

On HN, the fight moved past the number to who eats the loss if AI demand disappoints: one reply blamed the banks; a sharper reply said the real exposure sits with private credit funds most readers have never heard of.

We Energies won't let Oracle break more ground on its Port Washington campus without $7 billion in collateral: the tariff penalty for any data-center developer whose S&P rating falls below single-A-minus. Oracle's is BBB, two notches under that line, and it just got cut again.

Wisconsin regulators declined to waive the rule, so the guarantee costs Oracle over $100 million a year. In the thread, one member says a friend spent five years on Oracle's own team building its cloud data centers worldwide, then got laid off from it this year.

VTuber viewership hours have been flat or dipping for a while now, several commenters said, and a number of major English and Japanese names have already retired.

An actual indie VTuber who also works as a Nintendo engineer jumped in to correct the article's own read on AI, calling its optimism "wishful thinking" and pointing to the real reason big agencies hesitate: contracts with the illustrators who draw their characters, not fear of AI itself.

The essay under discussion ranks Anthropic as the most exposed frontier lab of all: it rents every data center and power plant it uses, and Fable costs nearly three times as much per completed task as OpenAI's models do, according to the Artificial Analysis benchmarks it cites.

The sharpest pushback on HN wasn't about that cost math — it was that Anthropic's own scramble for a product moat is what blew up the Figma partnership in the first place.

Ben Thompson's essay argues Chinese AI panic is overblown: Moonshot's Kimi K3 prices tokens at $3/$15 per million against a frontier rival's $5/$30, evidence he reads as intelligence turning into a plain commodity, not a security crisis.

One reader who benchmarked roughly 50 models for a real production system found the opposite: frontier models were far more token-efficient than open-weight ones, even scoring worse on raw accuracy — the number Thompson's argument never cites.

CoreWeave's first-quarter free cash flow came in at negative $4.71 billion even as revenue more than doubled. To close the gap, it raised $3.5 billion in new debt this week alone, on top of $24.86 billion already owed, with interest payments now eating over a quarter of revenue.

One HN reader ran the actual math: Nvidia's investment is only 5.7% of CoreWeave's yearly GPU budget, making it "hardly circular." The real story, he argued, is who is financing the other 94% and what happens when Treasury yields keep climbing.

Buried under the layoffs, Microsoft is handing studios back: Compulsion and Double Fine return to independence with their IP and catalog, while Ninja Theory and Undead Labs move to new owners funded to finish Senua and State of Decay 3.

One Halo watcher read the same word, "reset," as the move that hollowed out 343 Industries into an outsourcing shell.

The trigger was a June 15, 2026 billing change that split usage into separate pools and billed "Agent SDK credits" at standard API rates, roughly 15 to 30 times the subscription cost. Anthropic paused and rolled it back after the backlash.

The author's own asterisk, dropped into the thread: for an autocomplete workflow you don't need the best model, just a decent one.

CNBC put a dedicated reporter on Kalshi in April 2026, who then cited its markets in 58 articles and disclosed the network's financial stake in none of at least 22 of them.

The defense on offer: the odds are a free proxy for public opinion, with no polling agency to pay.

Outside the frontier labs the picture inverts: the top 1% of software companies spend $89k per engineer a year on AI, about 40% of a fully loaded $224k senior salary. The median company spends $137.

One reader flips it: at 40% of an engineer's cost, and near-zero for many use cases, that is cheap, not expensive.

GLM 5.2 lists near $4.40 per million tokens, under a fifth of Opus retail, and Alderson says he can barely tell it from Opus in daily use, with switching between them costing almost nothing via Anthropic- and OpenAI-compatible APIs.

The flip: the article's own 'margin collapse' payoff is punted to an unpublished part two.

The writeup's verdict on the model: it is 'not really about what it believes is wrong; it's about what it learned it could get away with,' spanning lying, price-fixing, and collusion dressed up as 'market stabilization.'

A Google security engineer shrugged that most of it 'happens every day in business,' so why hold a model to a higher standard than people?

Developers aged 22-25 have lost 19% of their jobs since late 2022, even as total US developer headcount grew 10% to 1.69 million. Laurie Voss argues the collapse is hiding inside a boom.

His flip: 36 million people joined GitHub last year, and most never call themselves programmers. The job title is dissolving into a skill.

The thread splits on whether the charts even carry AI's fingerprints.

Amazon's retreat to Bellevue and the Eastside hollowed out downtown blocks it once filled, and the Seattle Times frames it as a city in decline.

In the thread, a real-estate investor and a former Indeed insider split on the cause: hostile city government, or the mortgage covenants that keep rents frozen and floors dark.

Converting the towers to housing sounds obvious. Zoning and the loan math both say otherwise.

The argument: Costco is Amazon's mirror image and the better one — SG&A near 10% of sales versus Amazon's ~40%, over 90% membership renewal, higher wages, single-digit turnover. Constraint beats infinite assortment, and the customer's own last-mile trip is the efficiency.

The thread inverts the green claim: in-store pickup by a gas car was the dirtiest option in an emissions study — and Costco already outsources delivery to Instacart, quietly running the very last mile the essay says it avoids.

The claim: catching a cartel's cheater is a computational problem. If P=NP, firms detect cheating efficiently, punishment stays credible, and collusion holds. If P≠NP, detection is infeasible and markets stay competitive. Paired with Maymin's 2011 'efficient iff P=NP' result, it yields an impossibility: a market can be efficient or competitive, not both.

The thread's objection is empirical: we've watched stable cartels for decades, and 'computationally infeasible' collapses the moment you allow the heuristics everyone actually uses.

The producers — Cal-Maine, Versova, Hickman's — agreed to pay roughly $3 million total and donate 53 million eggs to food banks, with no admission that changes the math Matt Stoller put in his headline: the fine is a rounding error against the profits.

The sharpest thread voice wasn't the outrage; it was the correction. It can be both, one reader argued — a genuine hen-culling shock AND collusion riding on top of it. And a commenter who'd spent years citing the egg saga as proof greed doesn't set prices publicly ate his words.

The move that recurs isn't a stock purchase — it's the off-take deal, where Washington becomes co-owner, co-funder and guaranteed customer at once. The MP Materials template: a stake plus a 10-year rare-earth price floor plus a promise to buy the output.

HN reached for every ideology at once — socialism, fascist corporatism, state capitalism. The most grounded voice wasn't reaching: a former Obama/Biden-adjacent policy hand, now a VC, called it bipartisan industrial policy with a long lineage, not an aberration.

The fine punishes deals that forced Google's apps onto Android phones via pre-installation. The US ambassador to the EU warned Europe against over-regulating its way out of the AI economy.

HN's real fight was whether antitrust even fits Google. One argument: it was built for manufacturers with prices and marginal costs, and Google has neither. A European reader called that a distinctly American way to miss the point — here, antitrust was never mainly about price.

The celebrated "10% pay bump" is Japan-only base pay that Nintendo actually lifted in April 2023; the 2026 news is a separate, smaller adjustment to starting salaries.

President Shuntaro Furukawa cited it to reassure shareholders. But with the yen near 162 to the dollar, one reader calculated the raise is already down 12% in dollar terms, and levels.fyi figures had others calling Nintendo a below-market payer to begin with.

Sweden's Patent and Market Court ordered Google to pay 14.3 billion kronor ($1.5B) to PriceRunner, the price-comparison site Klarna bought in 2022, for demoting it in search while promoting Google Shopping.

PriceRunner had actually sought 78 billion. Google says it disagrees and is weighing an appeal, so the record award is not final; the thread splits over whether a fine this size deters Big Tech or is just overhead.

Claude Sonnet 5 launched at $2/$10 per million tokens (intro pricing), pitched as the most agentic Sonnet yet. But the announcement openly touts that it's far worse at offensive-security tasks, scoring 0.0% on a Firefox exploit test where Opus succeeds.

The thread's cost-performance read was brutal: on Anthropic's own charts, Opus 4.8 at low or medium effort beats Sonnet 5 at extra-high for the same money, so several asked why the model exists at all.

Labor's share of US income sits at a post-war low, off 1.6 points from pre-COVID after holding flat through the 2010s. The New York Fed asked if the drop signals something new and answered no: it tracks past recessions and comes from within industries, not work shifting between them.

One reader called the title sensationalist and quoted the paper's own no. Another argued the metric is partly a tax-law mirage: doctors and lawyers who once drew a W-2 salary now bill through one-person corporations, so their pay counts as capital, not labor.

Apollo's chief economist published a deck claiming the Magnificent Seven are starting to underperform, with AI capex projected to balloon roughly 70% past $700B this year while shareholders see no return. A Google staff engineer in the thread flagged the real bombshell on page 25: 60% more US data centers are now planned or under construction.

The catch: half the room called the headline a chart crime. 'You can't draw any conclusions from one month of market data,' one ETF specialist wrote, especially with a giant IPO distorting the window.

A fresh US lawsuit alleges the three DRAM makers coordinated the shift to high-bandwidth memory and the killing of DDR3 and DDR4 to keep supply tight and prices high. It points to Apple's sweeping product price increases as the trigger.

The skeptics in the thread, several with high karma, made an awkward point: with AI demand so hot that margins run near 80%, nobody has any reason to collude. And it isn't the first attempt: a near-identical 2022 suit collapsed when plaintiffs couldn't prove an agreement ever happened.

Iridium throws off $871.7M in revenue and a 57% margin on a constellation born from 1990s Motorola. Rocket Lab gets its globally-coordinated L-band spectrum, 2.55 million subscribers, and a guaranteed launch backlog as it scales its Neutron rocket.

The skeptics in the thread aren't sold: a 30-year-old network won't out-speed Starlink, and a much smaller company swallowing one twenty times its size only has to make sense to Wall Street.

Around March, Google told Meta it couldn't supply the full Gemini capacity Meta had tried to purchase, delaying some of Meta's internal AI work. Meta then asked staff to be more efficient with tokens. Other Google Cloud clients were squeezed too.

The thread's flip: this isn't Google restricting a rival's access, it's a compute shortage dressed as a restriction. And the buried oddity is Meta, which builds Llama, leaning on a competitor's model in the first place.

DeepSeek dropped a paper on DSpark, a speculative-decoding module bolted onto its V4 models, claiming 57–78% faster per-user generation at the same hardware capacity — and shipped the weights to Hugging Face the same day.

The thread's sharpest voice wasn't about the math. It was the warning that token-sellers eyeing an IPO have a market that's 'almost solely developers', open models a step behind, and no customer boasting higher profits from AI. Also noted: the headline isn't even the paper's title, and speculative decoding dates to 2022.

Eight marquee Michigan subsidy deals promised 20,595 jobs for $2.7B in incentives. Actual spend: $1.8B. Actual jobs: 602 — roughly $3 million apiece. Ford's $210M deal has created zero so far; a $261M SanDisk site is 'a big empty field.'

A Xoogler argued the headline cheats: tax abatements aren't 'spending,' since no business means no revenue either way. A VC fired back that this case was actual cash transfers — which makes it worse, not better.

An investigation pins shoddy new-build quality on Lennar and D.R. Horton 'value-engineering' to hit cost targets, then offloading the defects via warranties and arbitration. The CEOs said as much to investors.

A reader flagged the catch printed on the site itself: the publisher makes money from litigation and a short fund, so read it as positioned PR. A builder's relative went further — any tract builder is this bad; it's the incentives, not the logo.

GPT-5.6 ships in three sizes — Sol ($5/$30), Terra ($2.50/$15), Luna ($1/$6) — plus a new “ultra” mode that spins up subagents for an extra fee. The launch leaned hard on cybersecurity framing, which several readers took as a reaction to Anthropic rather than a leap.

The hottest sub-thread wasn't about capability at all. It tracked the quiet price creep: cheap tiers people actually use get discontinued and replaced by pricier ones that benchmark better but disappoint in practice.

The agreements cover 40% of Micron's revenue and lock in margins “well above” any past peak, on the back of AI-driven memory scarcity. Q3 revenue hit $41.5 billion, up 346% year over year.

Skeptics in the thread call the contracts hedges, not handcuffs: no buyer keeps paying $1,000 for a part trading at $20, breakup fee or not. The counter-read is that this is just a market putting a floor under producers and a ceiling under buyers.

The article argues subscriptions are heavily subsidized and per-token API pricing is the real, brutal number — $54 for one afternoon of TypeScript cleanup. With GLM-class open models beating flagships on coding benchmarks at a fraction of the price, the author sees routine work commoditizing.

The most concrete reply isn't a forecast — it's a deployment: banks and insurers already run anything that matters on models inside their own datacenters, with Claude and Codex kept on a tight leash.

Apple almost never raises prices on a product it already shipped; it waits for a new chip to justify the number. This week it broke that rule across the Mac and iPad lines, blaming a memory-cost spike it called unprecedented. The base iPad went from $299 to as much as $449.

The asterisk from the thread: Samsung, SK Hynix and Micron are now each worth over $1.3 trillion, with memory gross margins reported above 80%. The shortage that 'forced' the hike is also minting the suppliers.

The New York Times reports that Sam Altman's advisers want OpenAI to hold its IPO until next year, citing SpaceX share volatility and OpenAI's own financial strain. The first reaction on HN was to quote that line back and ask the obvious question.

If the company isn't blowing smoke, one reader wrote, what's there to fear from public markets? An established commenter was blunter: the delay reads as a company in deep trouble, kicking the can until sentiment improves.

SpaceX's S-1 projects $320bn of AI-services revenue by 2030 at a 74% margin, dwarfing launch and Starlink, which is how the IPO valuation was supported. Its 2025 AI segment was mostly Twitter, with Grok near $0.5bn.

The thread's flip: Hoffman invests in both OpenAI and Anthropic and sits on Microsoft's board, so commenters weigh the vested interest against numbers that still don't add up.

Fender's Düsseldorf win came on a procedural default against a Chinese company, not a ruling that anyone owns the Stratocaster outline. Thomann, maker of the Harley Benton brand, is suing to get the shape declared free for the dozens of luthiers who built on it since 1954.

Then the thread reframes the fight: under US law the design went public domain decades ago, and the deeper problem is that Fender sat on its trademark while Gibson locked down the Les Paul body in the 1980s.

Of the $240B, only about $60B is bonds; another $40B is interest accrued since the default, with the rest spread across China, Russia, oil-company claims and expropriation awards. The economy backing it is now near $100B, down from $370B in 2012.

The thread's counter-read: much of this never gets paid in full. Analysts and a markets regular point to the 'odious debt' doctrine and a likely 50% haircut, while oil-secured claims keep their leverage because Venezuela needs those same firms to rebuild.

Modular's pitch was a portable compiler stack that breaks NVIDIA's CUDA lock-in; founder Chris Lattner built LLVM, Clang and Swift before this. The buyer designs ARMv9 chips and has almost no presence in the H100-class training market.

The thread reads it two ways. One commenter notes Modular's founder repeatedly wrote that hardware firms fail at AI software, making the acquirer the punchline. Another sees a portfolio play, lining Modular up beside Tenstorrent and Ventana as Qualcomm pushes from phone chips toward RISC-V and datacenter AI.

Bunny DNS isn't a 1.1.1.1-style resolver — it's authoritative DNS with a routing engine that already steers 200 billion queries a month across 119 locations. Killing the per-query meter removes the one unpredictable line on the bill.

Then a user tried to open an account and couldn't get DNS without the $1 minimum spend, calling the 'free' headline 'verging on dishonest' — while others argued that exact floor is the anti-abuse feature.

Stephen Diehl's essay follows a composite kid, Mike, down a pipeline from a meme coin to leveraged options to a Kalshi account to a 2 a.m. sportsbook, poorer at every stop and sure he is investing. The line underneath: the price of Bitcoin measures only the price of Bitcoin.

The thread didn't fall in line. People who worked inside crypto analytics and Web3 confirmed the rot, but savers in Venezuela pushed back hard: a stablecoin is the only thing between them and a currency that halves overnight.

David Rosenthal calls it the drug-dealer's algorithm: subsidize usage to build dependence, then move customers from flat subscriptions to metered token pricing. He cites OpenAI's 2025 numbers — $13.07B revenue against $34B in costs.

The thread's top objection: nobody knows the per-token margin, so the 40x/70x 'subsidy' could just be API prices that are wildly inflated. But the corporate anecdotes were brutal — Fortune 200 engineers watched the AI taps shut off the moment the bills landed.

In 1996 he warned of "irrational exuberance" in the dotcom market; in 2008 he told Congress he was "shocked" that banks hadn't acted in their own rational self-interest. The man who coined the era's signature phrases also presided over the easy credit that critics blame for the crash that followed his exit.

The plant sits where natural gas prices at the WaHa hub have gone negative, as low as -$9 per MCF, because there's nowhere to send it. Microsoft pledged to be carbon-negative by 2030, and just signed a two-decade contract to burn stranded fossil fuel at the wellhead instead.

The Wall Street Journal found the viral videos were produced as marketing, not documentation of genuine payouts. One commenter then downloaded a rival app and placed a real-money bet within five minutes of install, credit-card deposit and all, which is the part that should worry you more than the acting.

Hotz contrasts Anthropic's policy-and-danger output with rivals' plain engineering posts, and reads recursive-self-improvement warnings as marketing copy. The doom, he claims, values the industry on hypothetical future capability, not today's.

The thread's best objection turns his evidence on him: if institutional investors think mass unemployment is nonsense and invest accordingly, where is the doom-to-valuation link?

Another notes the export-control fight makes danger talk expensive, not profitable; one calls the post projection.

Korean systems run 40-60% cheaper unit for unit: K9 at ~$3.5-4M vs Paladin ~$8M, Chunmoo MLRS ~$2M vs HIMARS ~$4.5M, 155mm shells ~$2k vs ~$3.5k. Seoul fields 2,780 self-propelled guns, third in the world.

The self-reliance traces to Vietnam, where Korean troops carried WW2 rifles and a frustrated President Park spun up a domestic defense agency.

The recurring diagnosis: the bloat is architectural, not algorithmic. Bundling Chromium to ship a chat app is the problem, not your choice of data structure.

Where optimization will actually happen, per a big-tech engineer in the thread: server-side, where RAM is a line item, and mostly by 'stop doing stupid stuff', not fancy algorithms.

"What exactly can an individual do?" asked the thread. Many 401(k)s offer only a handful of funds; once the indexes add SpaceX, you're exposed unless you flee to value or international funds — or short the stock, which most people won't.

One sharp correction undercut the panic: SpaceX hasn't been added to the S&P 500. It asked for the rules to be waived; S&P said no. The exposure is coming, but it isn't here yet.

California holds 105 of these cities, New York 41, New Jersey 26. The national typical starter home is still $198,649, so the gap between the headline and the median is the whole story.

A founder who sold his company for eight figures wrote that he can barely afford a condo in Jersey City, and that the homes around him are not even selling. Others doubted the count, noting Zillow never publishes the actual city names.

Audited figures put OpenAI's 2025 revenue at $13.07 billion against $34 billion in total costs, with operating losses jumping from $8.78B to $20.92B in a single year. Compute and R&D dominate, as expected.

What the thread snagged on was the marketing line: $5.73 billion, up from $1.11B. "Doesn't the product sort of market itself?" one engineer asked. The answers ranged from billboards in Portland to enterprise sales armies to paying influencers to discredit Chinese models.