THE AI BUBBLE – OR HOW THE MARKET IS LOSING ITS MIND IN REAL TIME
The AI market is at peak euphoria. Companies are signing multi-billion-dollar deals, swapping equity stakes, pumping capital – all under the banner of a “new technological era.”
But behind the flashy headlines and ever-growing numbers, the same old symptoms are reappearing – the ones we saw in the late ’90s, right before the dot-com bubble burst.
DEAL OF THE YEAR: AMD × OpenAI
On Monday, $AMD officially announced a multi-year partnership with Microsoft-backed #OpenAI to supply 6 GW of GPU capacity for future AI infrastructure.
The first wave will include 1 GW of Instinct MI450 chips, with rollout scheduled for the second half of 2026.
The key detail: warrants for $AMD stock.
OpenAI will have the right to buy 160 million $AMD shares (roughly 10% of the company) at a symbolic price of $0.10, provided several conditions are met:
▫️ #OpenAI must deploy infrastructure up to 6 GW of GPU capacity;
▫️ $AMD must hit both technical and commercial milestones;
▫️ one of the stages activates when $AMD reaches $600 per share.
In theory, it’s a perfect “win-win”: both sides are incentivized to make the partnership succeed.
In practice – it raises more questions than answers.
$NVDA, $AVGO, $MSFT, $GOOGL, $AMZN, $META, $ORCL, $NBIS, $CRWV, $EQIX, $CSCO
The AI market is at peak euphoria. Companies are signing multi-billion-dollar deals, swapping equity stakes, pumping capital – all under the banner of a “new technological era.”
But behind the flashy headlines and ever-growing numbers, the same old symptoms are reappearing – the ones we saw in the late ’90s, right before the dot-com bubble burst.
DEAL OF THE YEAR: AMD × OpenAI
On Monday, $AMD officially announced a multi-year partnership with Microsoft-backed #OpenAI to supply 6 GW of GPU capacity for future AI infrastructure.
The first wave will include 1 GW of Instinct MI450 chips, with rollout scheduled for the second half of 2026.
The key detail: warrants for $AMD stock.
OpenAI will have the right to buy 160 million $AMD shares (roughly 10% of the company) at a symbolic price of $0.10, provided several conditions are met:
▫️ #OpenAI must deploy infrastructure up to 6 GW of GPU capacity;
▫️ $AMD must hit both technical and commercial milestones;
▫️ one of the stages activates when $AMD reaches $600 per share.
In theory, it’s a perfect “win-win”: both sides are incentivized to make the partnership succeed.
In practice – it raises more questions than answers.
$NVDA, $AVGO, $MSFT, $GOOGL, $AMZN, $META, $ORCL, $NBIS, $CRWV, $EQIX, $CSCO
MUTUAL INCENTIVE OR MUTUAL INFLATION?
$AMD is effectively paying OpenAI with its own shares to buy its chips.
OpenAI, in turn, benefits from AMD’s stock going up – because its future stake depends on it.
That creates a feedback loop where both sides are inflating each other.
Wedbush analysts called the deal a “1996 moment” – the start of a new IT era.
A more accurate label might be a “2000 moment” – when #Cisco financed its own clients through venture arms so they could buy $CSCO gear.
We all know how that ended.
PARALLELS WITH NVIDIA
The $AMD-OpenAI story looks even more worrying given what $NVDA is doing.
$NVDA announced plans to invest up to $100 billion in #OpenAI to deploy 10 GW of AI infrastructure – again, built on its own GPUs.
Ironically, OpenAI could use part of those funds to buy chips from $AMD.
Meanwhile, its business remains unprofitable and continues to burn cash.
What we’re seeing is a “circle of capital”:
$NVDA invests in OpenAI → OpenAI rents compute from $ORCL → $ORCL buys chips from $NVDA.
A perfect ecosystem – until you remember what happens to such systems when new money stops flowing.
Image by Lin @Speculator_io
$AVGO, $MSFT, $GOOGL, $AMZN, $META, $NBIS, $CRWV, $EQIX
$AMD is effectively paying OpenAI with its own shares to buy its chips.
OpenAI, in turn, benefits from AMD’s stock going up – because its future stake depends on it.
That creates a feedback loop where both sides are inflating each other.
Wedbush analysts called the deal a “1996 moment” – the start of a new IT era.
A more accurate label might be a “2000 moment” – when #Cisco financed its own clients through venture arms so they could buy $CSCO gear.
We all know how that ended.
PARALLELS WITH NVIDIA
The $AMD-OpenAI story looks even more worrying given what $NVDA is doing.
$NVDA announced plans to invest up to $100 billion in #OpenAI to deploy 10 GW of AI infrastructure – again, built on its own GPUs.
Ironically, OpenAI could use part of those funds to buy chips from $AMD.
Meanwhile, its business remains unprofitable and continues to burn cash.
What we’re seeing is a “circle of capital”:
$NVDA invests in OpenAI → OpenAI rents compute from $ORCL → $ORCL buys chips from $NVDA.
A perfect ecosystem – until you remember what happens to such systems when new money stops flowing.
Image by Lin @Speculator_io
$AVGO, $MSFT, $GOOGL, $AMZN, $META, $NBIS, $CRWV, $EQIX
THE ROI PROBLEM
Greenlight Capital warns that “extreme” and “unprecedented” capital spending in tech could lead to massive value destruction.
The scale is staggering – and it’s unclear how these investments will ever pay off.
Examples:
▫️ $NVDA → $100B in #OpenAI
▫️ $NBIS × $MSFT → $17.4B over 5 years
▫️ $CRWV × $META → $14.2B over 6 years
▫️ $GOOGL Cloud → $25B to expand AI infrastructure within PJM grid (Pennsylvania)
▫️ $AMZN → Pennsylvania → $20B
▫️ $AMZN → Australia → AWS $13–14B over 4 years
▫️ $AMZN → North Carolina → $10B
▫️ $GOOGL → $6B in capex for 2026
▫️ South Korea (gov program) → $7.26B by 2026, incl. 15K GPUs and a national AI center
▫️ South Korea (investment fund) → $34.4B to support AI and semiconductors
▫️ HUMAIN (Saudi Arabia) → multi-billion project targeting 1.9 GW by 2030 and 6 GW by 2034
▫️ AWS × HUMAIN → $5B AI Zone project
▫️ $EQIX (Saudi Arabia) → $1B hyperscale data centers
▫️ In September, $AVGO announced a $10B order for AI chips from a new client, reportedly OpenAI
Total AI-infra investments now run around $367B per year, roughly 1.2% of global GDP – an insane number.
THE ECONOMICS DON’T ADD UP
Even the most optimistic projections say AI could add +15 p.p. to global GDP by 2035, or about +1% per year.
But since most of the equipment is imported, the real contribution to domestic GDP will be minimal – imports are subtracted from GDP.
TS Lombard analysts note that the “AI-driven growth boom” narrative sounds great but is deeply misleading.
Most of the money just circulates inside the tech sector.
$ORCL, $CSCO
Greenlight Capital warns that “extreme” and “unprecedented” capital spending in tech could lead to massive value destruction.
The scale is staggering – and it’s unclear how these investments will ever pay off.
Examples:
▫️ $NVDA → $100B in #OpenAI
▫️ $NBIS × $MSFT → $17.4B over 5 years
▫️ $CRWV × $META → $14.2B over 6 years
▫️ $GOOGL Cloud → $25B to expand AI infrastructure within PJM grid (Pennsylvania)
▫️ $AMZN → Pennsylvania → $20B
▫️ $AMZN → Australia → AWS $13–14B over 4 years
▫️ $AMZN → North Carolina → $10B
▫️ $GOOGL → $6B in capex for 2026
▫️ South Korea (gov program) → $7.26B by 2026, incl. 15K GPUs and a national AI center
▫️ South Korea (investment fund) → $34.4B to support AI and semiconductors
▫️ HUMAIN (Saudi Arabia) → multi-billion project targeting 1.9 GW by 2030 and 6 GW by 2034
▫️ AWS × HUMAIN → $5B AI Zone project
▫️ $EQIX (Saudi Arabia) → $1B hyperscale data centers
▫️ In September, $AVGO announced a $10B order for AI chips from a new client, reportedly OpenAI
Total AI-infra investments now run around $367B per year, roughly 1.2% of global GDP – an insane number.
THE ECONOMICS DON’T ADD UP
Even the most optimistic projections say AI could add +15 p.p. to global GDP by 2035, or about +1% per year.
But since most of the equipment is imported, the real contribution to domestic GDP will be minimal – imports are subtracted from GDP.
TS Lombard analysts note that the “AI-driven growth boom” narrative sounds great but is deeply misleading.
Most of the money just circulates inside the tech sector.
$ORCL, $CSCO
THE COST DYNAMICS
Back in February, Sam #Altman said “the cost of using a given AI level drops about 10× every 12 months.”
If that’s true, ROI on infrastructure spending will inevitably collapse.
AI access keeps getting cheaper, but GPUs, power, and maintenance don’t.
That imbalance could become the trigger for a future investment bust.
BOTTOM LINE
I side with Greenlight: it will be extremely hard to recoup these investments.
Right now, it all looks like a classic bubble – just wrapped in AI branding.
We’ve seen this movie before:
investors overestimate a technology, money spins inside the sector, the market feeds on itself – until new capital stops arriving.
The internet did change the world – but profits came 15-20 years later, not in 2000.
AI feels eerily similar.
Everyone wants in, everyone’s chasing the future – but tangible results are still missing.
The question isn’t whether the AI revolution will happen – it’s when, and how much collateral damage it will leave behind.
The bubble may have some air left in it, but this $AMD-#OpenAI deal looks like the beginning of the end.
The market is clearly losing its mind over AI.
No one knows how long the hype will last – a month, a quarter, or a year – but one thing’s for sure: the landing will hurt.
p.s. More on $ORCL coming Sunday. Stay tuned.
$NVDA, $AVGO, $MSFT, $GOOGL, $AMZN, $META, $NBIS, $CRWV, $EQIX, $CSCO
Back in February, Sam #Altman said “the cost of using a given AI level drops about 10× every 12 months.”
If that’s true, ROI on infrastructure spending will inevitably collapse.
AI access keeps getting cheaper, but GPUs, power, and maintenance don’t.
That imbalance could become the trigger for a future investment bust.
BOTTOM LINE
I side with Greenlight: it will be extremely hard to recoup these investments.
Right now, it all looks like a classic bubble – just wrapped in AI branding.
We’ve seen this movie before:
investors overestimate a technology, money spins inside the sector, the market feeds on itself – until new capital stops arriving.
The internet did change the world – but profits came 15-20 years later, not in 2000.
AI feels eerily similar.
Everyone wants in, everyone’s chasing the future – but tangible results are still missing.
The question isn’t whether the AI revolution will happen – it’s when, and how much collateral damage it will leave behind.
The bubble may have some air left in it, but this $AMD-#OpenAI deal looks like the beginning of the end.
The market is clearly losing its mind over AI.
No one knows how long the hype will last – a month, a quarter, or a year – but one thing’s for sure: the landing will hurt.
p.s. More on $ORCL coming Sunday. Stay tuned.
$NVDA, $AVGO, $MSFT, $GOOGL, $AMZN, $META, $NBIS, $CRWV, $EQIX, $CSCO
@InvestingVisual Just posted it! Tagging you like we agreed. Hope you like it.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
