DanielRomero

The amount of hindsight geniuses…
Now everyone was bearish on neoclouds
But nobody said anything when they were at all-time highs
They only turn bearish once they are already down 30%

$MU being down 10% on fears of compute oversupply is just funny
This market doesn’t have the slightest clue where AI is or where it is going
It’s just responding to any daily stimulus
Until we get hard data on the productivity improvements derived from AI it’s going to be lumpy

The concern is not just $META
xAI made the same pivot
When two of the biggest AI labs turn from pick-and-shovel buyers into sellers, the market starts fearing cyclicality
If selling compute is more profitable than consuming it internally, what that signals is the possibility of oversupply on the horizon
$META has spent aggressively on AI infrastructure, with more than $180B of commitments tied to future data center buildouts
The market now sees that those commitments may not be repeated. That is undoubtedly a blow to data center companies and neoclouds
Now, there is one important thing to con

The concern is not just $META
When two of the biggest AI labs turn from pick-and-shovel buyers into sellers, the market starts fearing cyclicality
If selling compute is more profitable than consuming it internally, what that signals is the possibility of oversupply on the horizon
$META has spent aggressively on AI infrastructure, with more than $180B of commitments tied to future data center buildouts
The market now sees that those commitments may not be repeated. That is undoubtedly a blow to data center companies and neoclouds
Now, there is one important thing to consider
If AI requires a h

Dylan Patel on Orbital Data Centers:
"If you look at 2040, I think probably more than half of incremental compute will be going in space. But if you look at 2030, I think it's sub 1%."

Jensen watching every semi company 10x while $NVDA is still below its October 2025 highs

MLCC prices are rising so fast Chinese traders are updating quotes every 30 minutes
TrendForce reports that in Shenzhen’s Huaqiangbei market, traders are reportedly updating MLCC quotes every 30 minutes as demand from AI servers, AI ASICs and EVs collides with tight high-end supply
Consumer electronics are still weak, but AI hardware needs far more high-capacitance, high-reliability MLCCs
That is pulling capacity toward premium specs and away from lower-end parts
Some high-end MLCC prices have reportedly risen several times this year, with Japanese and Korean suppliers in the strongest positio

On SemiAnalysis’ (not) recent report...
- $NVDA wanted a 4-die configuration with 16 HBM stacks inside one advanced package
- However, that is not possible at high yields yet without glass substrates
Instead, Rubin Ultra will move to a 2+2 configuration: two dual-die packages on the same board
The system still gets four GPU dies, and the Kyber server still preserves the intended compute and HBM footprint. The design is being split across the board because the package is not ready for volume
This is a workaround while glass substrates and PLP ramp through 2028
TSMC CoPoS mass production is stil

Ouster $OUST is now up:
+93% from my first buy
+148% since my update reinforcing conviction
At one point, I was down 40% on my original shares. Now, I’m up 137% overall
That is the power of building conviction and accumulating during weakness

Cantor Fitzgerald raised its $AMD price target from $500 to $700 while keeping an Overweight rating
The firm said AMD has the strongest compute momentum among semiconductor companies
Cantor still views $NVDA and $AVGO as inexpensive, while taking a more cautious view on $QCOM, arguing that the stock could revert to being valued mainly as a smartphone story

$AMD will probably announce Anthropic as a customer on July 22
From Citi a month ago:
“We believe AMD has won Anthropic as a customer for its MI450 AI accelerator, based on our discussions with industry contacts, and expect AMD to announce it at Advancing AI Day in July,” Citi analysts wrote

I had no idea it was this bad
$IREN seriously paid the most any company has ever paid to sponsor an American club
The Warriors’ previous Rakuten patch deal was $20M per year. IREN’s is paying more than 2.5x that
That is crazy.

Semi bottlenecks are now flowing into materials
Specialty gases are becoming the main pressure point
Tungsten hexafluoride, or WF6, is used to deposit tungsten films inside chips, especially in DRAM, NAND and advanced logic
Prices have surged more than 200%, partly due to China’s tungsten export controls and potential output cuts from Japanese suppliers Kanto Denka Kogyo (4047.T) and Central Glass (4044.T)
This benefits alternative suppliers like SK Specialty, Foosung (093370.KS), CSSC Specialty Gases ( Haohua Gas ( and Grandit ( while the customers paying higher prices include Samsung Electro

Physical AI memory demand will be much larger than people expect
$MU has said humanoid robots carry 10x the memory content of an average L2+ vehicle
The average car today has around 16GB of DRAM, while L4 autonomous vehicles can require over 300GB
Humanoid robots are expected to use compute platforms comparable to high-end autonomous vehicles
Given that, if physical AI scaled to 100M humanoid robots, or compute-equivalent robots, the total DRAM need would be:
- 100M x 300GB = 30EB
That is equivalent to around 75% of 2026 global DRAM capacity
We can already see proof of this with NVIDIA’s Jets

Jefferies expects memory prices to rise 40–50% QoQ in Q3 2026, followed by another 30–40% QoQ increase in Q4 2026
$MU is guiding for $31 EPS in calendar Q3 2026
If prices rise another 35% and then remain stable through 2027, Micron would be trading at less than 7x 2027 EPS
However, Jefferies is even more optimistic than that
For 2027, Jefferies expects memory pricing to remain 40–45% higher YoY. Relief is not expected until 2028, when new capacity could add 15–20% supply, although the article says this may not fully offset AI and compute demand
Realistically, Micron may be trading at less than

A $100B fully diluted market cap was insane
$CBRS is now coming back down to reality
For those interested...
If they achieve $6B in revenue by 2028 at their targeted 60% gross margin, that would imply $3.6B in gross profit
So now it’s trading at 10x 2028 estimated gross profit (optimistically) on a basic market cap basis. On a fully diluted basis, it’s 15x
Operating leverage will still be far from optimal by then, but $AMD and $AVGO trade at roughly 30x and 20x 2026 gross profit, respectively
So Cerebras isn’t necessarily insanely expensive anymore, but it has relatively weak market positionin

Memory investors talking with management teams for the past 6 months