Texas regulators are resuming environmental permits for data centers
“The agency has resumed its normal permit review process and will evaluate applications under all applicable state and federal laws and regulations,” TCEQ spokesperson Laura Lopez said in a written statement.
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me.
I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA.
I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation.
Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO.
Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours.
Jensen
Robotics foundation models could unlock a $150 billion market by 2036, according to @ABIresearch.
The reason is generalization: instead of being programmed for one repeatable task, robots running foundation models can increasingly adapt to new environments and unfamiliar objects. This extends automation to more complex, variable work in segments like manufacturing, warehousing, and healthcare.
But better models are only part of the picture. As ABI notes, the market’s growth also depends on the ecosystem around these models.
And a critical piece of that ecosystem is compute. Running these models on-device demands silicon with the real-time responsiveness and safety that physical systems require.
That’s why @Velaura_AI is building a purpose-built Physical AI SoC for robots, drones, and autonomous systems.
Read more from ABI Research on the growing Physical AI market opportunity: abiresearch.com/press/ai-pow…#physicalai#robotics#automation
Let me make it simple for everyone:
Jensen DID NOT secretly“smuggle” any $NVDA chips to Mainland China.
Nvidia sold chips to MegaSpeed a Chinese company that Trump, Lutnick and Bessant were all fully aware of years ago and allowed them to be sold to them.
Nvidia and Jensen did nothing wrong and was up front to the US Commerce department from day one. MegaSpeed is not on the Department of Commerce RESTRICTION list EVEN TODAY! 🤣
Why? Why did Trump and Lutnick allow them to be sold you ask?
1) Trump and Lutnick wanted rare earths from China and used these H200 gpu sales as a bargaining chip. We give you, you give us deal. They have what the US want, US has what China wants.
2) Trump added a 25% govt tax to all GPUs sold to China to benefit the US economy and the treasury’s coffers. Trump wanted bonus income into the treasury. He can kill 2 birds with one stone.
Understand now all?!?!?!??
Jensen however is currently being made a scapegoat 😩
I’m starting to give more credence to the conspiracy theory that Jensen turned a blind eye to some of Nvidia’s advanced chips being smuggled into China.
If he did, I think it was because he believed it was in Nvidia’s best interest. I don’t think Chinese spies got to him.
I don’t believe in coincidences, especially when it comes to $MARA.
Here’s an interesting connection worth looking into:
In 2024, Worldwide Mission Critical (WWMC) was involved in overseeing Starwood’s Renaissance Technology Park project in Virginia, where Amazon ($AMZN) was identified as the end customer.
Fast forward to 2026:
WWMC is now involved in MARA and Starwood’s Project Horizon in Kearney, Nebraska, and is also working on the massive Hannibal, Ohio data center development.
So we have:
Amazon → Starwood → WWMC → MARA’s AI infrastructure projects.
And remember what Fred Thiel recently said: Starwood has already worked with major hyperscalers, and revealing more details about one potential lease would make the tenant’s identity obvious.
Does this confirm Amazon is the mystery tenant? No.
But the same development team that worked on Amazon’s infrastructure is now involved in MARA’s projects, right as the company says it’s approaching major lease signings.
Maybe it’s just a coincidence. Personally, I don’t think so.
The verdict:
The plant isn't the story. The hedges are the lock.
Every MWh that rolls off a hedge reprices from ~$35 toward market.
Watch:
• Tap pricing vs par
• The hedge expiry schedule in the pro formas
• The closing 8-K
Clue #6: the 2032 wall.
Long Ridge notes: due 2032.
MARA's $950M 0% converts: due 2032.
@MARA has borrowed at 0%. Here it chose 8.75%.
Why? Secured at the asset, ring-fenced inside Long Ridge, zero dilution for $MARA holders.
Clue #5: why the gap exists.
PJM's 2027/28 capacity auction cleared at the $333/MW-day cap. Uncapped, it would have been ~$530.
The whole grid fell short of its reliability target.
Data centers are the demand. Long Ridge is supply sitting inside the shortage.
Clue #4: FTAI already gave us the formula.
In 2024: resetting hedges from $28 to $42 (+$14/MWh) ≈ +$50M EBITDA.
Same ratio on a ~$35 gap ≈ ~$125M a year.
On a plant MARA pitched at $144M EBITDA.
Napkin math off one winter quarter. But it shows where the lever is.
Clue #3: the scary line on the balance sheet.
Derivative liabilities: $310M, up $87M in 9 months.
Book equity: -$37.6M.
Bears will say "negative equity." It's not operating losses. It's hedge marks, and they grow when forward power prices rise.
Real cost until expiry (-$33M realized in 9M). Also a receipt for what each MW is now worth.
Clue #2: Long Ridge's own Q1'26 deck.
Realized merchant price: $70.23/MWh
Realized price incl. hedges: $35.40/MWh
A year earlier the gap was $13.
Now it's $35.
The hedges FTAI signed are eating half the price.
Clue #1: a number decided back in April.
The 8-K is a $185M add-on to Long Ridge's 8.75% 2032 notes, conditioned on the deal closing.
$600M existing + $185M new = $785M.
That's exactly the debt MARA said in April it would assume.
In May they paid ~$1.5M in consent fees to keep these notes alive. The plan was set from day one.
$MARA's new power plant sold electricity at $70/MWh last winter.
It kept $35.
Today's 8-K, which nobody read, shows where the other half goes and why it may be the most bullish line in the filing.
Case file 🧵🔍
Yesterday $MARA CEO Fred Thiel sat on a panel at Cantor Convergence called "Conversion is Inevitable: The Path to AI Infrastructure"
I went through every minute of it. Here are Fred's key quotes, and what I think he's telling us
◆ THE LEASES
Cantor asked straight up about the "two by year end" from the earnings call. Fred:
"The only reason we're sitting waiting for the tenant to sign the lease is, in a couple of cases, there's some internal delays"
"Sometimes it needs to go from one person's desk to another person's desk and that person has to sign it"
"Nothing permit-wise or power-wise"
"No technical issues"
Then the line that matters most:
"If I were to give more details, as regards one lease, it would become very evident who the tenant is"
Read that again. If the details alone would give the tenant away, this is a name you would recognize. Small startups don't have a paper trail that long between desks. Big organizations do
◆ THE PHONE IS RINGING
"Since the election we have gotten a lot more phone calls"
Anything with power and a clear path to permitting is "very interesting regardless of size"
MARA has power in hand. The buyers have to come to the people who have it
◆ WHY STARWOOD
Hyperscalers are "spending 90% fixing problems with first time developers"
"That's one of the reasons we did the Starwood deal. They've done deals with Google, Microsoft, Amazon. They've built 7 to 8 gigawatts of capacity for them. They operate it"
Nobody picks a partner with that track record to chase small tenants. You pick that partner because you're building for the biggest names in the room
◆ THE URGENCY
"There was a bit of an oh shit moment. People need capacity to come online next year"
Capacity next year means power this year. Power is exactly what MARA has
◆ KEARNEY, NEBRASKA: THE MATH
Fred called these numbers illustrative:
- 100 MW site, contributed at about $1 Million a MW
- Starwood matched with $100 Million
- $15 Million per MW build cost, $1.5 Billion total
- $300 Million of equity total. Each side has put in $100 Million and adds $50 Million more
"Talk about capital light"
A $1.5 Billion AI campus for roughly $150 Million of MARA cash. That's the setup the Street keeps missing
◆ THE RIOT COMPARISON
Asked about Riot's concrete-pad deal, Fred went right at the math:
"33% yield on a $6 million build cost is essentially the same as 15%, 17% yield on cost on a $12 million build cost, or what a hyperscaler is going to pay you on a $20 million per megawatt build cost"
"Step away from yield on cost and look at what the real NOI is"
Then he asked out loud whether Riot, with 900 MW at Corsicana, could have gotten a better deal from a hyperscaler
Translation: MARA isn't racing to sign the first lease offered. They're holding out for the deal that pays the most on what's actually earned
◆ EXAION
Built inside EDF to run the data centers at its nuclear sites. When EDF wouldn't put in more capital, MARA stepped in and recapitalized it
The first announced client is "a very sophisticated model builder" going from signing to online in "a few months," running prior-generation GPUs
"Good profitability, very attractive pricing... great relationships with both NVIDIA and other providers"
◆ THE CLOSE
"Demand is outstripping the capacity. Capacity can't come online fast enough"
Fred cited agentic token usage up eightfold since August
Then he said more companies are moving to open-weight models on their own private cloud, and that "the bare metal business has to convert to private cloud business"
Stop there for a second
Private cloud with its own orchestration layer is exactly what Fred said Exaion does. He may have just described where the whole industry is heading, and MARA already owns a piece of it
◆ PUTTING IT TOGETHER
Power in hand
A partner that builds for Google, Microsoft and Amazon
A tenant so recognizable he can't say more
A private cloud operator ready for where AI is going next
A balance sheet that's already cleaned up
Demand is outstripping capacity, and @MARA is sitting on the capacity
The first signature is the spark
NFA / DYOR