JY762

We need $MSFTB this afternoon; they have to carry the entire burden of AI trading on their own

$BEB After the recent short-selling report, it has delivered investors the strongest operational response they could possibly hope for, with revenue for the first time breaking through the $1 billion mark, product sales growing more than threefold, and management raising guidance across the board.

This momentum is driven by the value of bringing AI capacity online faster, as Bloom estimates that a 1 GW data center can generate up to $24 billion in annual revenue—meaning that saving one month can unlock roughly $2 billion in revenue that would otherwise be delayed.

So Bloom isn’t just selling power—it’s selling “time to token revenue.”

$SOFI fell more than 10% and broke below $15👀

Michael Burry says that the five-year credit default swaps for $NVDA are surging because the company is “overextending” to sustain recurring AI spending.

But that assumes vendor financing is creating demand, while Nvidia is still constrained by supply and is using financing to speed up deployments for customers who have already been waiting for chips.

$SOFI Five years ago:
• $15 per share
• 2.5 million members, with an annualized gross profit rate of $300 million

$SOFI Today:
• $15 per share
• 15.8 million members, with an annualized gross profit of $2.5 billion

The price is the same, but the nature of the business is completely different.

Wooden-Sister Cathie Wood today purchased $25 million worth of $SPCXB and $TSLAB

$SKHYB Second-quarter earnings

• Revenue: expected $54.5B vs. $57.7B
• Operating profit: $41.6B, expected to be $44.2B.
• DRAM average selling price: up 30% quarter over quarter
• NAND average selling price: up 55% quarter over quarter

Third-quarter outlook
• DRAM shipment volume: up 10% quarter over quarter
• NAND flash shipment volume: up 3% quarter over quarter

FY2026 outlook
• DRAM demand: up 25% year over year
• NAND demand: up 18% year over year
• Signed long-term supply agreements with 10 customers

$BE Q2 performance far exceeded expectations

• Revenue was $1.07B, above the expected $827M
• EPS was $0.78, above the expected $0.41
• EBITDA was $253M, expected to be $253M $149M
• Gross margin: 34% versus the expected 32%

Fiscal year 2026 guidance
• Revenue $4.05B vs. Est. $3.74B
• EPS was $2.70, above the expected $2.17

The Trump administration has proposed allowing the U.S. Federal Aviation Administration to waive certain environmental reviews for commercial launch and reentry vehicles.

This change could speed up the approval process.
$SPCXB and $RKLBB , in the next ten years, launch activities will increase by more than twofold.

$AMZN The report says that the usage of AI shopping assistants will “more than double” in 2025, with more than half of shoppers now willing to let AI handle the entire purchasing process.

Setting their own agents as the default agents for Amazon and Alexa helps protect their $72 billion advertising business from being undermined by third-party agents that bypass search and sponsored listings.

$SKHYB revenue is expected to grow from about $90 billion in 2027 to over four times that amount, reaching about $365 billion

This growth is driven by its HBM. SK hynix provides about 60% of the memory content for the $NVDAB Vera Rubin platform.

Global DRAM market share

Samsung ~39%
$SKHYB ~29%
$MUB ~22%
$CXMT~8%

Because Samsung, SK hynix, and Micron control 90% of the market, and as capacity shifts toward higher-margin HBM via take-or-pay contracts and price floors, the industry is more disciplined when it comes to pricing.

CXMT is a true commodity DRAM competitor. It is currently the highest market-cap listed company in China, with a valuation of $500 billion. However, due to the lack of meaningful HBM business, it has not yet moved into the highest-profit AI memory segment.

Still buying Nvidia? $MUB has just fallen below $800
Give me short!

$SNDKB fell more than 50% in the past month😳

$AMKR.US fell more than 20% 😳

Latest news: Apple $AAPL market value has first broken through the $5 trillion mark.

Moonshot AI is seeking more $NVDAB Blackwell GPUs to train Kimi K4. It expects its scale will exceed the 2.8 trillion-parameter K3, despite facing U.S. export restrictions.

K3 has been used on Blackwell compute by multiple Chinese cloud service providers and exhausted its inference capacity within 48 hours.

$AMDB profit could grow 10x in the next three years, as AI capacity promised by $META, OpenAI, and Anthropic grows alongside the expansion of the $ORCL
50,000 GPU MI450 supercluster.

This demand-driven setup powers two profit engines: Instinct GPUs drive AI growth, while EPYC Venice expands the share of high-margin servers, and Helios racks capture more value from each deployment.

The biggest upside may come from inference, where 31 TB of HBM4 per rack and agentic workloads push the CPU-to-GPU ratio toward 1:1, potentially accelerating both Instinct and EPYC at the same time.

$TER.US Q2 preliminary earnings report

• Revenue: $250 million vs. expected $174 million
• EBITDA: (-$13 million) vs. expected $2 million
• Module production: 835 MW
• Cash: $79 million

T1 increased its G2 Austin Phase 1 cost estimate by 20% to $510 million and pushed initial production back to Q1 2027, but still expects 2026 production to be near the top end of its 3.1 GW to 4.2 GW range.

$TSLAB signed a long-term agreement to purchase 90% of the output from a 509 MW solar and 360 MW battery project in Arizona.

The KKR-backed facility is expected to come online in 2028 as growth in AI data centers continues to tighten the U.S. power market.