Goldman says 95% of enterprises are getting zero return on $...

TFTC

npub1sk7mtp67zy7uex2f3dr5vdjynzpwu9dpc7q4f2c8cpjmguee6eeq56jraw

hex

6b97b204660a3cbc4209554bedbfbc9fff0ef3bc463df0b6f8788b7305b4248b

nevent

nevent1qqsxh9ajq3nq509uggy42jldh77fllcw7w7yv00skmu83zmnqk6zfzcprpmhxue69uhhyetvv9ujuem4d36kwatvw5hx6mm9qgsgt0d4sa0pz0wvn9yck36xxezf3qhwzksu0q254vruqed5wvuavuswf3pnr

Kind-1 (TextNote)

2026-05-06T13:46:10Z

Goldman says 95% of enterprises are getting zero return on $30-40B in GenAI spending.

Goldman's own Exhibit 9 shows consumer AI demand growing 12x by 2030.

Both are true. Here's what it means:

Enterprise AI is failing because legacy companies don't know how to use these tools. They're spending millions bolting chatbots onto workflows designed in 2005 and wondering why ROI is zero.

The real AI wave is startups who build from scratch. Three people doing the work of thirty. No legacy systems. No "AI transformation consultants." Just raw efficiency gains from day one.

Meanwhile, the demand side is undeniable. Token economics turn positive H1 2026. Consumer agents are shifting from chat sessions to always-on. The hockey stick is already forming.

The question is who captures the value. Open-source LLMs and agent frameworks are commoditizing inference. If small businesses and individuals can route around hyperscalers, the value moves from chips to applications. The market is overpricing hyperscaler equity as if they have traditional big tech moats. They might not.

And the backstop nobody's talking about: Trump invoked the Defense Production Act for AI. The federal government is treating this as critical infrastructure. Will there be misallocation? Yes. But a full bust is hard to sell when demand outstrips supply and the government won't let it fail.

Enterprise AI is a bubble. Consumer AI is a boom. Goldman proved both in the same report. https://blossom.primal.net/41b737e010ccc33b20ff58b91c18f8021b8a6a555f18475b28eb9891e2c6ad5f.jpg

原始 JSON

{
  "kind": 1,
  "id": "6b97b204660a3cbc4209554bedbfbc9fff0ef3bc463df0b6f8788b7305b4248b",
  "pubkey": "85bdb5875e113dcc99498b474636449882ee15a1c78154ab07c065b47339d672",
  "created_at": 1778075170,
  "tags": [
    [
      "imeta",
      "url https://blossom.primal.net/41b737e010ccc33b20ff58b91c18f8021b8a6a555f18475b28eb9891e2c6ad5f.jpg",
      "m image/jpeg",
      "x 41b737e010ccc33b20ff58b91c18f8021b8a6a555f18475b28eb9891e2c6ad5f",
      "size 83974"
    ]
  ],
  "content": "Goldman says 95% of enterprises are getting zero return on $30-40B in GenAI spending.\n\nGoldman's own Exhibit 9 shows consumer AI demand growing 12x by 2030.\n\nBoth are true. Here's what it means:\n\nEnterprise AI is failing because legacy companies don't know how to use these tools. They're spending millions bolting chatbots onto workflows designed in 2005 and wondering why ROI is zero.\n\nThe real AI wave is startups who build from scratch. Three people doing the work of thirty. No legacy systems. No \"AI transformation consultants.\" Just raw efficiency gains from day one.\n\nMeanwhile, the demand side is undeniable. Token economics turn positive H1 2026. Consumer agents are shifting from chat sessions to always-on. The hockey stick is already forming.\n\nThe question is who captures the value. Open-source LLMs and agent frameworks are commoditizing inference. If small businesses and individuals can route around hyperscalers, the value moves from chips to applications. The market is overpricing hyperscaler equity as if they have traditional big tech moats. They might not.\n\nAnd the backstop nobody's talking about: Trump invoked the Defense Production Act for AI. The federal government is treating this as critical infrastructure. Will there be misallocation? Yes. But a full bust is hard to sell when demand outstrips supply and the government won't let it fail.\n\nEnterprise AI is a bubble. Consumer AI is a boom. Goldman proved both in the same report.\nhttps://blossom.primal.net/41b737e010ccc33b20ff58b91c18f8021b8a6a555f18475b28eb9891e2c6ad5f.jpg",
  "sig": "d8e4c8c9424e8db11cde8283fe88814ec1578bba389c238e0f2b38be43fd4471588d77d15beeb4fc2767a9b36a3397da49a8960c3f7e10e5e9d4c7a30e7f67aa"
}