Knowledge Vault 7 /373 - xHubAI 02/09/2025
🫧BURBUJA.AI ¿A punto de explotar?
< Resume Image >
Link to InterviewOriginal xHubAI Video

Concept Graph, Resume & KeyIdeas using Moonshot Kimi K2 0905:

graph LR classDef bubble fill:#ffd4d4,font-weight:bold,font-size:14px; classDef gov fill:#d4ffd4,font-weight:bold,font-size:14px; classDef hype fill:#d4d4ff,font-weight:bold,font-size:14px; classDef china fill:#ffffd4,font-weight:bold,font-size:14px; classDef europe fill:#ffd4ff,font-weight:bold,font-size:14px; classDef labour fill:#d4ffff,font-weight:bold,font-size:14px; classDef infra fill:#f9f9d4,font-weight:bold,font-size:14px; Main[Vault7-373] Main --> B1[95 % pilots
zero value 1] B1 -.-> G1[Bubble] Main --> B2[Data governance
not models 2] B2 -.-> G1 Main --> H1[Hype plateau
then lift-off 3] H1 -.-> G2[Hype] Main --> H2[AI hijacked
marketing circus 4] H2 -.-> G2 Main --> H3[AI winter
after hype 5] H3 -.-> G2 Main --> H4[OpenAI PER
>400 6] H4 -.-> G2 Main --> H5[Gartner peak
inflated 7] H5 -.-> G2 Main --> P1[Email pilots
lack depth 8] P1 -.-> G1 Main --> P2[€4 k intern
fails fast 9] P2 -.-> G1 Main --> P3[No ROI
95 % fail 10] P3 -.-> G1 Main --> C1[China decades
vs quarters 11] C1 -.-> G3[China] Main --> C2[Open China
threat incumbents 12] C2 -.-> G3 Main --> E1[Europe lost
general models 13] E1 -.-> G4[Europe] Main --> E2[SME gap
blocks scale 14] E2 -.-> G4 Main --> L1[Junior roles
already gone 15] L1 -.-> G5[Labour] Main --> L2[Reskill to
orchestrate agents 16] L2 -.-> G5 Main --> I1[On-prem GPUs
beat cloud 17] I1 -.-> G6[Infra] Main --> I2[Data centers
guzzle water 18] I2 -.-> G6 Main --> I3[CAPEX arms
race billions 19] I3 -.-> G6 Main --> I4[ROI now
vs years 20] I4 -.-> G6 G1[Bubble] --> B1 G1 --> B2 G1 --> P1 G1 --> P2 G1 --> P3 G2[Hype] --> H1 G2 --> H2 G2 --> H3 G2 --> H4 G2 --> H5 G3[China] --> C1 G3 --> C2 G4[Europe] --> E1 G4 --> E2 G5[Labour] --> L1 G5 --> L2 G6[Infra] --> I1 G6 --> I2 G6 --> I3 G6 --> I4 class B1,B2,P1,P2,P3 bubble class H1,H2,H3,H4,H5 hype class C1,C2 china class E1,E2 europe class L1,L2 labour class I1,I2,I3,I4 infra

Resume:

The debate, moderated by Plácido Domenech, opens with gratitude and a warning: the summer’s anti-AI headlines are not just click-bait but symptoms of a deeper tension between market expectations and technical reality. A recent MIT report claiming that 95 % of generative-AI pilots return zero value is used as a springboard to ask whether we are inside a financial bubble, a technical winter, or a paradigm shift that incumbent firms simply misunderstand. Eduardo Cano, digital-transformation consultant, argues that the bubble narrative is premature; what looks like failure is rather the plateau between hype waves, similar to the early-2000s dot-com pause, while the underlying technology keeps maturing. Álvaro González, investor and biotech entrepreneur, counters that business narratives have hijacked AI discourse, turning a scientific revolution into a circus of press releases; he fears capital is being poured into storytelling instead of durable infrastructure. Antonio Flores Galea, university professor and author, reminds the panel that every AI hype cycle ends when short-term investors withdraw, but the real indicator is whether models are becoming embedded in production pipelines, which they are, albeit slowly.
The conversation quickly moves from finance to engineering rigor. José, a data-centric engineer, calculates that OpenAI’s implicit PER exceeds 400, a valuation reminiscent of the most speculative dot-com stocks, and warns that the gap between promised AGI and delivered arithmetic is widening. Ignacy, corporate trainer, stresses that most firms are stuck at the “Peak of Inflated Expectations” on the Gartner curve: they adopt Copilots for email generation yet skip the harder work of cleaning data, redesigning processes, and training staff. Several participants share war-stories of €4,000 pilots where interns fine-tune Llama on dirty CSV files and declare defeat when ROI does not appear in a quarter. The panel agrees that the 95 % failure statistic is less a verdict on AI than on corporate readiness: absence of data governance, absence of clear KPIs, and absence of technical literacy among executives who treat foundation models as plug-and-play silver bullets.
Looking forward, the debate converges on three take-aways. First, the real bubble is in financial expectations, not in the technology itself; China’s long-horizon state investments and open-source model proliferation will outlast American quarterly-capital cycles. Second, Europe has already lost the general-purpose model race but could still dominate industrial AI if it leverages SAP-like incumbents and IoT data; yet without strategic patience and SME financing, the continent risks becoming a digital colony. Third, the labour impact is neither myth nor mass extinction: AI is already displacing junior content and coding tasks, but the bottleneck is re-skilling workers to orchestrate agents rather than prompting them like chatbots. The session closes with an appeal for humility: learn the engineering, read books not headlines, and treat AI as infrastructure—electricity, not magic.

Key Ideas:

1.- MIT report says 95 % of gen-AI pilots yield zero value, sparking bubble fears.

2.- Panel agrees failure reflects poor data governance, not model limits.

3.- Eduardo sees hype-cycle plateau; historical analogies predict eventual lift-off.

4.- Álvaro warns business narrative hijacked AI into marketing circus.

5.- Antonio cites past AI winters triggered by investor withdrawal after hype.

6.- José calculates OpenAI implicit PER >400, exceeding dot-com extremes.

7.- Ignacy places most firms at Gartner “Peak of Inflated Expectations.”

8.- Email-generation pilots dominate but lack segmentation or profiling depth.

9.- €4,000 intern-driven Llama fine-tunes declared failure within quarters.

10.- Absence of ROI metrics and process redesign drives 95 % failure rate.

11.- China’s state-backed decades-long AI strategy contrasts US quarterly capitalism.

12.- Open-source models from China threaten closed US incumbents.

13.- Europe deemed lost for general models but could lead industrial IoT AI.

14.- SME financing gap blocks European AI adoption at scale.

15.- Labour impact: junior content & coding roles already displaced.

16.- Re-skilling bottleneck: workers must orchestrate agents, not prompt chatbots.

17.- Security fears push firms toward on-prem GPUs despite cloud economies.

18.- Data-center water consumption emerges as hidden environmental cost.

19.- CAPEX arms-race: Google $85 bn, Amazon $118 bn, Meta $70 bn.

20.- Short-term ROI demands clash with multi-year infrastructure payback.

Interviews by Plácido Doménech Espí & Guests - Knowledge Vault built byDavid Vivancos 2025