Nvidia's B200 now resells for 58% above list — and lenders noticed

A year-old AI accelerator is appreciating like real estate, not like a graphics card. That changes what gets borrowed against it — and what it says about the shape of the boom.

Nvidia's B200 is trading at 158% of its launch price roughly a year after it reached mass availability, according to Silicon Data's analysis — an accelerator that added 58% instead of shedding the 20% to 40% that standard depreciation schedules assume. Older A100s and H100s are also running well above what three- and five-year straight-line schedules would imply. The cause is scarcity, not sentiment: HBM3E memory is short, fresh allocations are expensive and slow to land, and buyers keep bidding for units already sitting in racks rather than waiting in line. The unit economics back it up — the B200 runs R1-class inference at roughly $0.20 per million tokens and 76 tokens per second per user, so the asset keeps earning while it ages.

Why it matters: an asset that holds its value and produces metered, billable output is exactly what lenders look for, so GPU fleets are increasingly financed against their resale value the way aircraft are, instead of written down like IT gear. That is the same logic behind Nvidia backs $500B data center deal with GPU value guarantee, and it cuts both ways. Used hardware priced at or above new hardware erases the margin advantage secondhand buyers used to count on, and CFOs lose the write-downs they once planned around. The one thing that could break the curve is a fast HBM4 ramp, and TrendForce does not see memory relief arriving before late 2027.

Moonshot AI launched Kimi Finance, a packaged financial-industry stack, and named the institutions already running it: ICBC, CITIC Construction Investment, CICC and E Fund, with dozens more in the programme. The pitch is specific — more than 10 authoritative data sources wired in through MCP (Wind, East Money, S&P Global, Cailianshe, Caixin Data), nine finance skills, five compliance and safety measures, and institutional-grade modeling and report delivery. The nine skills are named workflows, not vague capability claims: institutional financial modeling, research reports, institutional decks, financial-chart generation, earnings commentary, consensus maps, portfolio review, position morning briefs and an HK IPO lens.

The pilot numbers are the part worth reading twice. With CITIC Construction Investment, Moonshot built a trustee-report agent that has run across more than 50 issuers and 100-plus outstanding corporate bonds, producing over 60 reports; per-report handling time fell from about 30 minutes to 10, and human input dropped 67%. Integration work that was scoped at two months finished in three working days. Elsewhere the company reports financial modeling dropping from five to seven person-days down to half a day to one, and industry deep-research cycles compressing from 10–20 days to about two.

Why it matters: a model vendor shipping a vertical stack turns the data terminal from a prerequisite into a competitor, and the incumbents' moat was always distribution rather than the models. Financial institutions are also the most reluctant buyers of anything that can leak data or hallucinate a number, so a named list of banks and brokerages is a bigger signal than any benchmark score.

What to watch: whether B200 residuals survive the HBM4 ramp, and whether Kimi's finance deals convert into disclosed production seats rather than pilot logos.

If the hardware keeps gaining value, should the labs be buying compute or borrowing against it? Tell us in the comments.

AI 101 — How do AI video generators work?

An AI video generator is a model that starts with a grid of random noise and repeatedly refines it into a moving picture that matches your prompt — the same trick image generators use, stretched across time. Nothing is filmed, and nothing is stitched together from clips the model has seen. The pixels in the finished video were invented, frame by frame, all at once. Why it matters right now Kling 4.0, Kuaishou's flagship video model, went fully live on October 15 with up to 30 seconds of vid

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Read more

AI 101 — How do AI video generators work?

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Brain organoids classified Braille at 83% — in a trio, not alone