Michael Burry’s current holdings

[ICUMD]

Probably early innings with Nvidia.  However,  Googles TPUs seem to be able to do a phenomenonal job at a much lower cost. They are also now starting to sell their chips. 

 

Further, to Burrys point, many of the companies using NVIDIA chips will eventually be non survivors.  

 

So NVIDIAs torrid growth is not guaranteed down the line. 

 

It's anyone's guess when the music will stop.   

[winjitsu]

 

Seems accurate. Bulltard (options trader), which is above Burry, has been on record stating he has a little >$1mm ARR. 

Still, I believe Burry could over-take Doomberg. Same theme an audience and doomerism always sells

[This2ShallPass]

6 hours ago, Gregmal said:

Eh, 50x with growth isnt a big deal. People said it for like 15 years with the likes of Netflix, Costco and stuff like Amazon, and now get emotional W's outta "stock hasn't done anything for a couple years" or "its off double digits from ATHs" but the truth is that many things have printed people a ton of money while carrying PE's like that. It's way more relevant to focus on where you think they are in the growth cycle. NVDA could still be kinda early in the cycle. 

For sure, many of us missed the biggest bull run due to "valuation".  I just worry when you reach the stratospheric multiple trillion valuations, these high P/Es are not justified.

 

I think Palantir is the one that is a much surer bet, selling to the government should never have a 200x FCF multiple. Inherently unpredictable, subject to political winds etc.

 

[This2ShallPass]

Also, Burry's core point is still valid. GPUs are using a 4-6 year depreciation but Nvidia releases a new chip every year. One of these has to give, even assuming a Training->Inference->Maintenance / legacy waterfall for chip usage.

Hyper scalers are increasingly offloading data center risk to Neoclouds. Neoclouds get their debt using the credit worthiness of their customer (Microsoft / Nebius & Coreweave) and Nvidia is invested in them (I'm sure Nvidia will help out if these companies have trouble borrowing). Nebius expects $8B revenue next year but has to spend $30B capex (80% of it is GPUs)! For ref, they had $117M revenue last year and $350M TTM. And that capex is only 2026, in 2027 they'll have to repeat all of this and some more to buy the latest GPUs. The sheer scale and circular nature of all these deals seem trouble to me.

[kh812000]

I think there is way too much ink written on the "circular financing" narrative.  The only unique point of these relationships is just the sheer size and scale.  As an example, there is so much fud around the ORCL CDS which rose to 100 bps. But why is there no detail or press on what the implied default probability is for such a CDS rate beyond the "scare" headline??  Its because this implies a 1.6% probability of default!!  Not scary at all....

 

Now, is it really circular financing? NO. Its not at all like the dot com days...  For example, the OAI and NVDA Deal:  Cash ultimately originates from external capital (eg venture) and AI customers, which flows to OAI then to NVDA for real revenues in exchange for real equipment, which is needed to support the real customer demand.  This is economically different from a fake loop where a company secretly funds a customer purely to book cosmetic revenues w/o business value.  

 

This AI doom and gloom is all fud, and the key thing to watch is token demand / revenue generation by the AI cos.  As of now, its exponential growth and demand, both from growing usage, agentic AI, increasing intelligence/use cases.  Once token demand stops growing exponentially, then there is potentially overbuild worries, which likely will happen someday but not for a few years.

 

 The narrative of accelerated depreciation is also pure FUD.  This is written and discussed by many others...

[TwoCitiesCapital]

On 11/25/2025 at 5:09 PM, DegenerateGambler said:

Why would he short NVDA when it's like 50 PE growing 30%+ per year?  He has to make sure this is the very top of the bubble to make money.   Costco is like 50 PE growing less than 15% per year, he should be shorting that one instead if he wants to short 50 PE megacaps.  I know growth rate and PE is not the end all be all but just back of the napkin calculations tells you NVDA is not extremely expensive and the AI boom better be at the very top for him to make money on it.  Unlike 2008, the companies he picked to short are not crap companies - just highly valued, especially Palantir.  I think Palantir is more likely to work out for him than NVDA, but picking on NVDA is kind of ridiculous as a short thesis.   Worst case growth slows to 20% and NVDA falls to 150 and his puts still are underwater due to time premium burn.

 

I believe part of his thesis is that NVDA is over earning today. When you look at the circular financing of NVDA investing in OpenAI, and OpenAI turning that cash right back over to NVDA for chips or signing contract with Oracle so they can buy the chips from NVDA, it begs the question what is actual chip demand if OpenAI hits an air pocket? 

 

So 50x an over-earning company who may see earnings fall by 30-40% instead of grow 30-40% isn't crazy to me. The timing is the tough part. 

 

 

On 11/25/2025 at 7:58 PM, adesigar said:

 

Because. am I’m probably wrong on this. He’s compared Nvidia to Cisco during the .com bubble. My guess is thinks that AI is over hyped and the build out is over done. In a year or two the AI use cases won’t exist and there will be too much infrastructure and sales will fall off a cliff.

+1 

 

 

On 11/25/2025 at 1:10 PM, gfp said:

It's kind of cringe-worthy really.  He's extremely unlikely to be harping on this narrative at the exact top in AI stocks.  Much more likely that we are far enough away from a blow-off top that people reading his (I'm sure well reasoned) arguments and acting on them are going to get killed.

 

I've never been great at timing - so take any of my opinions with a grain of salt.

 

But I think it's telling that the Coreweaves, Supermicros, and even Oracles of the world are well off their tops. Either this is a significant lull in the bull market narrative despite NVDA continuing to blow out its earnings or the speculative names in the market are signaling the shift in sentiment. If the latter? The top is already in for most of the names - NVDA included - and the market is leading the deterioration in their financial results. 

[gfp]

36 minutes ago, TwoCitiesCapital said:

I've never been great at timing - so take any of my opinions with a grain of salt.

 

But I think it's telling that the Coreweaves, Supermicros, and even Oracles of the world are well off their tops. Either this is a significant lull in the bull market narrative despite NVDA continuing to blow out its earnings or the speculative names in the market are signaling the shift in sentiment. If the latter? The top is already in for most of the names - NVDA included - and the market is leading the deterioration in their financial results. 

 

It's just a shift in leadership.  The big spenders are being rotated out of and I assume the energy, pharma, and economically sensitive stuff is going to take a turn.  Small caps will start to perform better relative to the mega cap momentum names that previously led.  Doesn't mean the AI bubble is finished inflating.  On to the next bottleneck.  If it ain't GPUs it's power or something else.

[DegenerateGambler]

8 hours ago, gfp said:

 

It's just a shift in leadership.  The big spenders are being rotated out of and I assume the energy, pharma, and economically sensitive stuff is going to take a turn.  Small caps will start to perform better relative to the mega cap momentum names that previously led.  Doesn't mean the AI bubble is finished inflating.  On to the next bottleneck.  If it ain't GPUs it's power or something else.

Bloom Energy (BE) went from 8 to 147 on the AI power supply narrative.  It's crazy.  NRG a traditional power company literally tripled already in the last year on the AI power supply theme.  So you are correct that power is the bottleneck since Microsoft literally have GPUs collecting dust in their inventory since they cannot find enough power, but the market has front-ran this thesis already.

[DegenerateGambler]

9 hours ago, TwoCitiesCapital said:

 

I believe part of his thesis is that NVDA is over earning today. When you look at the circular financing of NVDA investing in OpenAI, and OpenAI turning that cash right back over to NVDA for chips or signing contract with Oracle so they can buy the chips from NVDA, it begs the question what is actual chip demand if OpenAI hits an air pocket? 

 

So 50x an over-earning company who may see earnings fall by 30-40% instead of grow 30-40% isn't crazy to me. The timing is the tough part. 

 

 

+1 

 

 

 

I've never been great at timing - so take any of my opinions with a grain of salt.

 

But I think it's telling that the Coreweaves, Supermicros, and even Oracles of the world are well off their tops. Either this is a significant lull in the bull market narrative despite NVDA continuing to blow out its earnings or the speculative names in the market are signaling the shift in sentiment. If the latter? The top is already in for most of the names - NVDA included - and the market is leading the deterioration in their financial results. 

The problem with his NVDA short is simply that there are so much better shorts out there (Oracle and Coreweave etc) and he would need a bear market to trigger some kind of big waterfall for puts to print.  I get shorting Palantir, not NVDA.  Also Oracle was a much safer short because how the heck is OpenAI with 20 billion dollars of revenue going to pay Oracle 350 billion in the next 5-10 years?  and Oracle was projecting their datacenter revenue to grow to 144 billion in 5 years, yeah right.  Then Oracle has to spend all the money building all this infrastructure HOPING some company will use their service for all their capital investment.  NVDA on the other hand, is a pants-and-shovels company.  They are the guys selling the pants-and-shovels during this gold rush.  Why would he pick that one?  
I get the Cisco analogy tho, NVDA will for sure fall the way of Cisco some day, but using puts, he put a time limit on it.  Also if NVDA sells off 30%, the rest of the AI stuff will absolutely get crushed, just short those.

I feel like he wants to be Cassandra more than he wants to make money.

[This2ShallPass]

2 hours ago, DegenerateGambler said:

The problem with his NVDA short is simply that there are so much better shorts out there (Oracle and Coreweave etc) and he would need a bear market to trigger some kind of big waterfall for puts to print. 

I cannot figure out the economics for these neoclouds (and extrapolating for the whole AI sector). Does anyone else see a way for this to work out, what am I missing?

I'll focus on Nebius since I looked into it. 

Expected connected power (EO26) = 900mW mid point (800-1G)
Expected connected power (EO25) = 220MW

Capex needed for 2026 = $30B (80% GPU, 1% land / contracted power, 19% datacenter buildout)

Capex for GPUs = $24B

Revenue for 2026 = $8B (mid-point)

If these GPUs can be used for 5 years - this is a big IF and core of Burry's argument. But let's say they provide same economic value for 5 years.

5 year revenue from today's Capex = $40B

Even w 5% interest, that's $38B cost of capex to get $40B revenue (Gemini says WACC for these high risk companies is 12-15% and 5-year NPV is -$12B).

Again, just a crude high level analysis using optimistic assumptions - simply cannot work.

Now add in all the actual hairy stuff:

1. Depreciation is 4 or 3 years

2. Capex is upfront and revenue over 5 years. Not guaranteed, need to add a discount factor for that.
2. Their cost of borrowing blows up
3. For 2027, they need to do this all over again...might need to borrow another $50B. If the neoclouds and hyperscalers don't / can't do this, then Nvidia revenue takes a huge hit in 2027.

 

[hasilp89]

https://omny.fm/shows/against-the-rules-with-michael-lewis/michael-burry-speaks

[Luke]

I am thinking of joining him in the palantir shorting...Nvidia is very very expensive and end market demand is not coming in close to what is priced in but if i look at Palantir its even worse. PE of what, 400? 430b company with 1b in earnings??

 

Anyone else short already? 

[Spekulatius]

On 12/1/2025 at 10:01 PM, DegenerateGambler said:

The problem with his NVDA short is simply that there are so much better shorts out there (Oracle and Coreweave etc) and he would need a bear market to trigger some kind of big waterfall for puts to print.  I get shorting Palantir, not NVDA.  Also Oracle was a much safer short because how the heck is OpenAI with 20 billion dollars of revenue going to pay Oracle 350 billion in the next 5-10 years?  and Oracle was projecting their datacenter revenue to grow to 144 billion in 5 years, yeah right.  Then Oracle has to spend all the money building all this infrastructure HOPING some company will use their service for all their capital investment.  NVDA on the other hand, is a pants-and-shovels company.  They are the guys selling the pants-and-shovels during this gold rush.  Why would he pick that one?  
I get the Cisco analogy tho, NVDA will for sure fall the way of Cisco some day, but using puts, he put a time limit on it.  Also if NVDA sells off 30%, the rest of the AI stuff will absolutely get crushed, just short those.

I feel like he wants to be Cassandra more than he wants to make money.

I agree  much better short than NVDA. PLTR (ridicolous valuation and politically vulnerable ), MU (commodity chip producer) , ORCL (tied to OpenAI, negative FCF), Corweave (also tied to OpenAI, Capex heavy) seem much better candidates.

[Sweet]

PLTR is an obvious short and the problem is that, based on the pricing, the options market also knows it’s an obvious short.

 

Edited December 6, 2025 by Sweet

[Luke]

42 minutes ago, Spekulatius said:

I agree  much better short than NVDA. PLTR (ridicolous valuation and politically vulnerable ), MU (commodity chip producer) , ORCL (tied to OpenAI, negative FCF), Corweave (also tied to OpenAI, Capex heavy) seem much better candidates.

Micron just rug pulled consumer products away from the market, i think the memory market really is developing in a oligopoly after all closely together with nvidia and other hyperscalers...at least these risks are there and there is so much money flowing into this that i dont like to short it. nvidia is also not as much an obvious short as you said than something like palantir...but still...tesla still going so hard after all the bad news...the apes are hard to shake off...

 

19 minutes ago, Sweet said:

PLTR is an obvious short and the problem is that, based on the pricing, the options market also knows it’s an obvious short.

 

yeah...on the other hand this thing can trade sideways or go up over periods of a year or more because people are lunatics and switched off their brains :))

[Marco Van Basten]

4 hours ago, Spekulatius said:

I agree  much better short than NVDA. PLTR (ridicolous valuation and politically vulnerable ), MU (commodity chip producer) , ORCL (tied to OpenAI, negative FCF), Corweave (also tied to OpenAI, Capex heavy) seem much better candidates.

Spek, where do you see PLTR's revenues and EBIT in 2035  and why?  Thank you.

[villainx]

6 hours ago, Luke said:

Micron just rug pulled consumer products away from the market, i think the memory market really is developing in a oligopoly after all closely together with nvidia and other hyperscalers...at least these risks are there and there is so much money flowing into this that i dont like to short it.

 

Sounds like you think MU might have room to keep going higher?  

 

My friend has been telling me to get into MU for the past three months, saying they turned the corner.  I'm not that into cyclical stuff like MU though.

[Luke]

13 hours ago, villainx said:

 

Sounds like you think MU might have room to keep going higher?  

 

My friend has been telling me to get into MU for the past three months, saying they turned the corner.  I'm not that into cyclical stuff like MU though.

Id stay away at these valuations. It was a great buy at 50 USD. 

[DegenerateGambler]

On 12/6/2025 at 2:04 AM, Luke said:

I am thinking of joining him in the palantir shorting...Nvidia is very very expensive and end market demand is not coming in close to what is priced in but if i look at Palantir its even worse. PE of what, 400? 430b company with 1b in earnings??

 

Anyone else short already? 

I am short Palantir since 188, but fed is lowering rates soon and QT ended and this bull-mania might continue for a while.  Shorting has the problem of requiring good timing, unlike the long side of the trade.

[Luke]

4 minutes ago, DegenerateGambler said:

I am short Palantir since 188, but fed is lowering rates soon and QT ended and this bull-mania might continue for a while.  Shorting has the problem of requiring good timing, unlike the long side of the trade.

nice. you are just short the shares right? no option games anything? since we are small scale trader we have patience and can pay the borrowing fees etc and sit it out. I think we are a price level that makes me comfortable holding short.

[DegenerateGambler]

35 minutes ago, Luke said:

nice. you are just short the shares right? no option games anything? since we are small scale trader we have patience and can pay the borrowing fees etc and sit it out. I think we are a price level that makes me comfortable holding short.

am actually short PLTU - the daily 2x palantir shares which mathematically moves in our favor as a short candidate over the long run

[KPO]

On 12/2/2025 at 11:56 AM, hasilp89 said:

https://omny.fm/shows/against-the-rules-with-michael-lewis/michael-burry-speaks

Thanks for sharing. One of the more honest investment interviews I’ve heard since Munger’s last.  His comments about the Fed and interest rates towards the end are spot on. What a rational person. Would be nice to hear from him once a year. 

[hasilp89]

14 hours ago, KPO said:

Thanks for sharing. One of the more honest investment interviews I’ve heard since Munger’s last.  His comments about the Fed and interest rates towards the end are spot on. What a rational person. Would be nice to hear from him once a year. 

np, i really enjoyed it as well. had never heard him talk before. think the media / movie / twitter distorted my view of him. more personable than i expected and very thoughtful

[Luke]

1 hour ago, hasilp89 said:

np, i really enjoyed it as well. had never heard him talk before. think the media / movie / twitter distorted my view of him. more personable than i expected and very thoughtful

burry is a great guy, smart guy...

[This2ShallPass]

22 hours ago, DegenerateGambler said:

am actually short PLTU - the daily 2x palantir shares which mathematically moves in our favor as a short candidate over the long run

You might already know this, but the 2x or 3x shares are only for short time periods. Volatility drag will eat into returns over the long term (anything over a month), even if you're right..

https://www.etf.com/sections/etf-basics/why-do-leveraged-etfs-decay

As per Gemini, the only exception where holding longer works:
 

This is the only time holding >1 month works. If the stock moves in one direction with very little volatility (e.g., goes up 1% every single day for 30 days), the daily rebalancing actually helps you. You will end up with more than 200% of the return.

• Why: You are compounding on a larger base every day.

• The Risk: This rarely happens. Stocks rarely go straight up without pullbacks. The moment a pullback happens, the volatility drag kicks in.