Listen to this Post
I’m looking for…
Home
My Network
Jobs
Messaging
23
Notifications
Me
For Business
Learning
Premium
Tony Moukbel
Tony Moukbel
Multi-Talented Innovator | 13 Innovations & 4 Patents | Cyber Security Expert | IT & Ai Engineering | 58 Certifications in Cybersecurity, Forensics, Programming & Electronics Dev.
Lebanon
UNDERCODE TESTING
Profile viewers
49
Post impressions
12
Your Premium features
Feed post
View Stephen Klein’s profile
Stephen Klein
• 2nd
Founder & CEO, Curiouser.AI | UC Berkeley Instructor | Reflective AI – Technology That Helps People Think | LinkedIn Top Voice in AI
View my services
1d •
Follow
NVIDIA, the leading seller of shovels in the AI gold rush, just admitted there is and has never been any gold.
Bryan Catanzaro, NVIDIA’s VP of Applied Deep Learning, told Axios last week: “For my team, the cost of compute is far beyond the costs of the employees.”
The story didn’t get much attention (and I think we all know why)
Three implications, all load-bearing for the entire AI investment thesis:
1. AI is cheaper than humans. It isn’t.
2. AI can replace humans. It can’t.
- The current AI frenzy is sustainable. It isn’t.
One Proof point: Uber’s CTO has already spent his entire 2026 AI budget. We are 25% into the year. 100% of the money, in 33% of the time. He is, in his words, “going back to the drawing board.”
1. AI is not cheaper than humans.
→ Swan AI: a $113,000 Anthropic bill for a four-person team. One month. $28K per person. Multiples of their salary.
→ Jensen Huang now says NVIDIA engineers earning $500K should spend $250K a year on tokens. Half their salary. To stay productive.
The pitch was labor arbitrage. The reality is labor plus compute.
2. AI cannot replace humans.
→ MIT’s 2025 NANDA study: 95% of enterprise GenAI pilots delivered zero measurable P&L impact — despite $30–$40 billion in spend.
→ Goldman Sachs chief economist Jan Hatzius: AI added “basically zero” to U.S. GDP growth in 2025. Of the 2.2% growth that year, ~0.2% was attributable to AI. The rest leaked offshore as imported hardware.
→ Tech laid off ~80,000 people in Q1 2026. Roughly half attributed to AI.
The layoffs were the easy part. The savings were the story. The story was wrong.
3. The frenzy is not sustainable.
A trillion dollars of capex is moving in circles. NVIDIA invests in OpenAI. OpenAI buys NVIDIA chips. Oracle hosts. Microsoft books revenue. Every dollar counted three times.
OpenAI is missing internal revenue and user targets. The Microsoft exclusivity deal just ended. Markets are wobbling. (
And how crazy is this? (When they miss revenue targets, that’s actually better for their business because what they don’t disclose openly is their revenue costs them money; there unit economics are all twisted and they have inverse economies of scale)
This was never really about productivity.
It was about narrative.
Replace humans → save money → look modern → stock goes up.
This has been a historic theatrical performance that will go down in the annals of business as truly extraordinary.
The question nobody on the AI 1.0 stage is willing to ask:
If the goal had never been replacement, if the goal had been to make the humans we already have radically more capable, would we be in this mess?
Curiouser.AI has products in the market today that augment human capacity and help users transcend their potential. We call it AI 2.0. Our goal is to drive innovation and create jobs. Visit us at (link in comments) to learn more.
The trick with technology is to avoid spreading darkness at the speed of light.
(Sources in comments.)
View image
77
46
View Stephen Klein’s profile
Stephen Klein
Author
Founder & CEO, Curiouser.AI | UC Berkeley Instructor | Reflective AI – Technology That Helps People Think | LinkedIn Top Voice in AI
1d
https://curiouser.ai/
4 reactions4
Reply
View Stephen Klein’s profile
Stephen Klein
Author
Founder & CEO, Curiouser.AI | UC Berkeley Instructor | Reflective AI – Technology That Helps People Think | LinkedIn Top Voice in AI
1d
Sources:
Catanzaro / NVIDIA on compute vs. employee cost — Axios, April 26, 2026
Uber CTO 2026 budget exhaustion — The Information, April 2026
Swan AI $113K Anthropic bill — Amos Bar-Joseph, LinkedIn / Tom’s Hardware
Huang on $250K/yr per engineer in tokens — Business Insider, March 2026
95% AI project failure rate — MIT NANDA, The GenAI Divide: State of AI in Business 2025
“Basically zero” GDP impact — Jan Hatzius, Goldman Sachs / Atlantic Council, Feb 2026
Goolsbee remarks — Federal Reserve Bank of Chicago, March 2026
Q1 2026 tech layoffs — Tom’s Hardware, April 2026
OpenAI missing targets, Microsoft deal end — The Information / Tom’s Hardware, April 2026… more
11 reactions11
Reply
1
View Joe Temple’s profile, open to work
Joe Temple, Open to work Verified Profile 3rd+Joe Temple • 3rd+
Adjunct Professor at Coastal Carolina UNiversity and Retired IBM Distinguished Engineer
7h
Stephen Klein spot on.
1 reaction1
Reply
View David Elfanbaum’s profile
David Elfanbaum Verified Profile 3rd+David Elfanbaum • 3rd+
AI generalist focused on the intersection of artificial intelligence, psychology and McLuhan-inspired media theory. Semi-retired.
1d
“You wrote, The trick with technology is to avoid spreading darkness at the speed of light.”
I think you’re dangerously close to doing with the opposite position what you’re rightly pointing to in the accelerationist camp.
“AI is cheaper than humans. It isn’t.” It definitely is cheaper in some cases. Like tier one customer support.
“AI can replace humans. It can’t.” It can in some cases, as in the above example.
3. “The current AI frenzy is sustainable. It isn’t.” I agree that frenzies by their nature aren’t sustainable. But I think it’s quite possible that the massive investments around AI infrastructure will continue for the foreseeable future.
… more
8 reactions8
Reply
5
See previous replies
View Sean Turner’s profile
Sean Turner Premium Profile 3rd+Sean Turner • 3rd+
AI-Driven Enterprise Revenue Leader | Architect of AI & Agentic Systems | SaaS, Industrial, & Digital Transformation | Builder of RAG & Conversational AI Solutions
1m
Ken C. I just asked ChatGPT about the current Senate leader and it said it was Chuck Schumer😂
Reply
View Pathaksa Tongpitak’s profile
Pathaksa Tongpitak 3rd+Pathaksa Tongpitak • 3rd+
🚀 SuperAffiliator @ AffiliateWeapons.com
1d
Spending billions to increase token consumptions without optimization. I think there is plenty of room for improvement, it’s just not talked about enough
3 reactions3
Reply
1
View Stephen Klein’s profile
Stephen Klein
Author
Founder & CEO, Curiouser.AI | UC Berkeley Instructor | Reflective AI – Technology That Helps People Think | LinkedIn Top Voice in AI
1d
Pathaksa Tongpitak Its an entirely new perspective that is required…it’s not a tech challenge, it’s “what problem are they REALLY trying to solve?” And can they solve it?
4 reactions4
Reply
View Jacob James-Vogel’s profile, open to work
Jacob James-Vogel, Open to work Premium Profile 3rd+Jacob James-Vogel • 3rd+
Independent Investment Analyst; Consultant @ Achieve By Design; Editor/ Creator @ X Gen Investor
1d
Thanks for this. Helpful as always. Lately I’ve been thinking about how the supply shortage in helium will impact companies like Nvidia. 33% of global production is off-line (Qatar) and 18% is unavailable to the west (Russia). That’s over half of global production now unavailable to the fabs in Taiwan and South Korea.
Supply chains are tricky- liquid helium is transported in specialized tankers at extremely cold temperatures and it vaporizes during transport. The longer the journey the less helium will be delivered.
The U.S. is the largest producer so fabs here should be fine- but who knows how much supply TSMC et. al have over in Asia. 45 days? 90 days? Couldn’t be much more than that.
With helium critical and irreplaceable in semiconductor fabrication it’s only a matter of time before companies like Nvidia are impacted by this shortage. … more
2 reactions2
Reply
1
View Stephen Klein’s profile
Stephen Klein
Author
Founder & CEO, Curiouser.AI | UC Berkeley Instructor | Reflective AI – Technology That Helps People Think | LinkedIn Top Voice in AI
1d
Jacob James-Vogel Wow I had no idea. That is really fascinating.
Reply
View Tikka Nagi’s profile
Tikka Nagi Verified Profile 3rd+Tikka Nagi • 3rd+
Sekura Ai I helping tech companies ship secure products without slowing down engineering. 20+ yrs | F5, IBM, Imperva | Built security at Gliimpse (acquired by Apple)
1d
It is a mistake to conflate a lack of immediate consumer AI breakthroughs with a lack of fundamental enterprise value. The real utility is quietly compounding right now by autonomously scaling workflows previously constrained by human bandwidth. In domains like security, where AI can continuously penetration test complex infrastructure at a scale manual teams simply cannot match, the ROI is already obvious.… more
1 reaction1
Reply
1
View Stephen Klein’s profile
Stephen Klein
Author
Founder & CEO, Curiouser.AI | UC Berkeley Instructor | Reflective AI – Technology That Helps People Think | LinkedIn Top Voice in AI
1d
Tikka Nagi DIsagree with you but we will see. Will Gen AI become a huge benefit to business? Absolutely. But is fear and a herd menatlity making CEOs stupid and reactionary? Look at the data. But anyway your view will be what they will all say after the crash
Reply
View Chris C.’s profile
Chris C. Premium Profile 3rd+Chris C. • 3rd+
Founder & Chief Coherence Architect, Samirac Partners LLC | AI Execution Control Layer | Admissibility at the Execution Boundary | Preventing Invalid Actions Before They Occur | Patent-Pending Architect
(edited) 1d
Interesting
One GPT call replacing 10 minutes of work → cheap
Running an entire enterprise workflow through LLMs → very expensive
So a more precise version is:
AI doesn’t reduce cost by default.
It reduces cost only when tightly
scoped and controlled.
Reality:
It replaces tasks
It reconfigures roles
It doesn’t replace systems of accountability
NVIDIA is absolutely printing money.
NVIDIA isn’t selling shovels to a fake market—they’re selling to the most capital-intensive buildout we’ve seen in decades.
What’s actually appears to be happening:
The “gold” isn’t labor replacement.
The “gold” is control of computation + infrastructure + execution paths.
–… more
2 reactions2
Reply
View Joe Glick’s profile
Joe Glick 2ndJoe Glick • 2nd
Biomimetic AI Pioneer, Chief Innovation Officer, Co-Founder, Polymath
1d
There exists a significant gap between the marketing hype of tech giants and the scientific reality published by their internal researchers, such as:
APPLE: The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity (doi 10.48550/arXiv.2506.06941)
MICROSOFT: The Illusion of Readiness in Health AI (arXiv 2509.18234)
GOOGLE: The Abstraction Fallacy: Why AI Can Simulate But Not Instantiate Consciousness
What is surprising is not the gap, but the scale of investment and the pressure to adopt without evidence of value and despite computing costs being multiples of savings from layoffs. Nvidia’s own VP of applied deep learning admits compute costs now dwarf employee costs. Healthcare ignores the mounting evidence of errors and risks to patents. Why?
When I was leading a metaontology project for the Nanoinformatics Society, a distinguished professor of physics and chemistry said, “we used to teach students how to make the world better with chemistry, but now we try to teach them how to undo the damage.” The AI hype is promising to fix everything with superhuman intelligence, picturing a quick and easy fix.
The expression “easier said than done” has never been more true.… more
2 reactions2
Reply
1
View Dragana Radić’s profile, open to work
Dragana Radić, Open to work Verified Profile 3rd+Dragana Radić • 3rd+
Digital/AI Transformation | Operational efficiency | Business value Delivery
13h
Joe Glick I would say this is liberal capitalism at its worse. Chasing fast and huge profit makes investing go into mode of extremely risky betting, when the times get rough. And when I say rough, I mean it in a ‘liberal capitalism’ way which is anything that yields less then 10% per year is considered not good investment.
1 reaction1
Reply
View Jeff Mike’s profile
Jeff Mike Premium Profile 3rd+Jeff Mike • 3rd+
Quantifying uncertainty in hiring and selection.
1d
Just restating for effect: The goal has always been replacement. The investor class lost control of the labor market for a brief time in 2022 and the resulting trauma will resonate across work for a decade or more.
1 reaction1
Reply
2
See previous replies
View Jeff Mike’s profile
Jeff Mike Premium Profile 3rd+Jeff Mike • 3rd+
Quantifying uncertainty in hiring and selection.
(edited) 1d
I hope we see the day…
Reply
View JK Fidden’s profile
JK Fidden 2ndJK Fidden • 2nd
Producer of the World’s 1st AI Film | Building the Future of Production | CEO & Founder | Father of Four
1d
OK so you do know the future is local, and not token gated?
I’m connected to some folk at the very cutting edge. AI development is exponential, and hardware limitations, cloud token gated systems, and massive data centers are now- they are not what’s inbound- and sooner than you may think. This really is AI 1.0 as you call it. 5.5 is being developed now, opensource. … more
2 reactions2
Reply
1
View Eduard Ruzga’s profile
Eduard Ruzga Verified Profile 3rd+Eduard Ruzga • 3rd+
CEO & Co-Founder at Desktop Commander
(edited) 1d
JK Fidden yeah. We are in pre personal computer, mainframe era of AI.
I think gpt 5.5 and opus 4.6 brought us into era of useful ai mainframes. If trends continue we will get local version by 2028, that will be personal computer era of AI. For now we are not yet there.
1 reaction1
Reply
About
Accessibility
Help Center
Ad Choices
Advertising
Get the LinkedIn app
More
LinkedIn Corporation © 2026
Tony Moukbel
MessagingYou are on the messaging overlay. Press enter to open the list of conversations.
Compose message
You are on the messaging overlay. Press enter to open the list of conversations.
Page inboxes
Click to see affiliated inboxes
▶️ Related Video (98% Match):
🎯Let’s Practice For Free:
IT/Security Reporter URL:
Reported By: – Hackers Feeds
Extra Hub: Undercode MoN
Basic Verification: Pass ✅
