Biznesom AI też się kończy eldorado

Pozwolę sobie wrzucić tutaj dwa niezwykle ciekawe artykuły, które wpadły w moje ręce dzięki @don_draper

https://ethanding.substack.com/p/ai-subscriptions-get-short-squeezed
https://ethanding.substack.com/p/windsurf-gets-margin-called

wlaśnie oglądałem film theo o tym
wyglada na to że niektórzy w nielimitowanym planie potrafią wygenerować koszty rzędu $3k dziennie.

BTW myślę że to już powszechnie znany trick, ale może ktoś nie słyszał - jeśli ktoś chce oszczędzić na tokenach to można przetłumaczyć input na chiński i potem odpowiedź z chińskiego. Po chińsku można zaoszczędzić 20-30% tokenów zachowując tę samą treść.

Researchers said “95 per cent of organisations are getting zero return” from their investments in generative AI, the technology that has sent US stocks soaring to record highs in recent months.

US tech stocks hit by concerns over future of AI boom

Mnie się ten boom na AI kojarzy z boomem na mobile w latach 2015-2020.
Oczywiście - nie neguję tutaj przydatności i potęgi tych modeli, jednak mam wrażenie że tak jak dzisiaj każda firma już ma swoją appke mobilna (a w 2016 roku miał ją też każdy grill i szczoteczka do zębów) tak za pare lat każda firma (poważna / produktowa) bedzie miala swoj model i beda go utrzymywac tak jak dzisiaj appki mobilne.

A tego całego syfu który się narobi obok nie jest mi szkoda, bo ja np. już szykuję rączki do sprzątanka bo vibe-coderach :)

myślę, że eldorado też skończy się eldorado.

Latest OpenAI numbers from the FT:

800m users, 5% paying (40m).
$13bn in ARR.
Implies a $325 annual ARPU, or $27/month per paying user.
70% of rev from subscriptions, rest is API.
$8bn loss in H1, prob $20bn run rate loss now? So basically spending $3 for each $1 in revenue.

https://on.ft.com/4nTP9GJ

Amerykańce mają kapitał to mogą latami sobie wrzucać pieniądze w taki biznes, ale większość tego wyścigu jednak nie przeżyje, szczególnie jak chce sprzedawać tylko tokeny z modeli, których nie posiadaja:

https://ethanding.substack.com/p/agents-vs-clouds

Tuta jeszcze fajne podsumowanie:

Odgrzebię temat:
Co sądzicie?

igor820 napisał(a):

Odgrzebię temat:
Co sądzicie?

Kto tym firmom splaci te zobowiązania jak my wszyscy będziemy bezrobotni przez AI ?

Może się nie kończy, bo kasa jeszce płynie, ale zmierza do j*bnięcia, IMO; acha to w temacie polecam też gary marcus substack.

W sumie to tutaj bardziej pasuje.

In the past, the CIO could argue that a new tech was in its early development, evolving, or "still being optimized," Dataiku's report notes, but regulators are looking for traceability, corporate boards are asking for performance, and investors want to see quicker value generation.

Perhaps not surprisingly, almost all respondents (98 percent) said pressure from the board to demonstrate measurable return on investment (ROI) is increasing, and retaining budgets may depend on whether they can prove a measurable return.

Time is running out, the report claims. Some 71 percent of the CIOs surveyed believe their AI budget will likely face cuts or a freeze if targets are not met by the end of the first half of 2026. And the consequences won't just stop at funding: 85 percent of CIOs believe their employers will tie their compensation to measurable AI outcomes, and many say the same applies to their chief exec.

CIOs also fret about the explainability of their AI systems, with some (29 percent) indicating they were asked multiple times over the past year to justify an AI outcome they could not fully explain.

Dataiku says this issue is about to harden into formal requirements, with 70 percent of CIOs expecting AI audit or explainability requirements to arrive within the next 12 months.

When it comes to the latest buzzword, "agentic AI" or AI agents with the power to carry out actions, the CIOs say these are no longer a lab concept and are already working inside corporate networks. Some 62 percent of CIOs indicate agents are embedded in some business-critical workflows, while a worrying 25 percent say agents have become the operational backbone of many critical workflows.

Worrying because the report reveals three-quarters of CIOs admit to not having full real-time visibility into AI agents running in production systems, even though they have the ability to implement actions.

The majority of CIOs (82 percent) admit their employees are creating AI agents and apps faster than the IT department is able to govern them. The same number indicate they are concerned worker-built AI could expose sensitive company data.

The fear is not only that employees will build the wrong thing – it's that they will build the right thing in the wrong place, with the wrong data, under the wrong controls, the report states.

Many CIOs are also owning up to a certain level of buyer's remorse when it comes to some of the decisions made regarding their organization's AI stack. 74 percent say they regret at least one major AI vendor or platform selection made in the past 18 months.

And while the execs in charge of the tech megacorps pushing AI are adamant the craze is most definitely not a bubble – except for OpenAI CEO Sam Altman, of course – the corporate CIOs in the survey are not so sure, and most are fearful of what will happen if it bursts.

Some 73 percent suspect their company would experience major disruption, with more than half (57 percent) saying their company's very survival might be at stake. Oh, and 60 percent feared they may lose their job should the big pop ever happen.

https://www.theregister.com/2026/02/17/no_roi_no_ai/