Chris

Thinking about intrinsic, the irresistible value proposition found in great disruptive businesses and next big thing and wave in the context of global politics.

[Chris @IntrinsicThink]

What exactly is an intrinsic?

It is the value proposition in a new business model that is irresistible to the consumer and surpasses all known modes of provision.

[NoLimit @NoLimitGains]

Jensen Huang, CEO of $NVDA, just highlighted $NOW, $CRWD, and $PLTR.

He said AI agents will not disrupt these markets. They will create the biggest opportunity these companies have ever seen.

[0xNobler @CryptoNobler]

🚨 CHINA JUST BROKE THE GLOBAL MARKET!!

The People’s Bank of China released new macro data today, and it’s far worse than expected.

Their economy is collapsing, and it'll drag global markets with it.

We’re not talking billions.

We’re talking TRILLIONS being pumped into the system to delay the collapse.

This is the kind of liquidity shock that breaks markets.

And it could spark the largest commodity squeeze the world has ever seen.

Here’s the real story no one is connecting yet:

China has officially launched the largest money-printing and credit expansion event in its history.

Their M2 money supply has gone vertical - now over $48 TRILLION (USD equivalent).

Pause and think about that.

That’s more than DOUBLE the entire U.S. M2 supply.

And unlike the West, China doesn’t print money just to levitate stocks.

Historically, when China expands money supply at this scale, it leaks straight into the real economy.

Hard assets.
Strategic resources.
Commodities.

China is exchanging freshly printed paper for REAL, limited-supply stuff:
→ Gold
→ Silver
→ Copper
→ Energy
→ Industrial metals

China is stockpiling metals, ramping domestic production, importing aggressively, and quietly reducing reliance on the Western financial system.

They’re also sitting on massive foreign assets: treasuries, equities, reserves - assets they can dump or weaponize if geopolitical tensions escalate.

This isn’t just economics.
This is financial warfare.

China, the world’s largest commodity buyer, is flooding the system with liquidity and hoarding hard assets.
Western banks are doing the exact opposite.
They’re reportedly sitting on enormous gold and silver short positions.

Roughly 4.4 BILLION ounces short.

Let that sink in.

Annual global silver mine supply?
~800 million ounces.

That means these institutions are short ~550% of the world’s yearly silver production.

Yes.
Five. Hundred. Fifty. Percent.

This isn’t “risky.”

It’s mathematically impossible to cover.

You cannot buy what doesn’t exist.

So here’s the collision course:

→ China debases its currency through relentless money printing
→ Liquidity spills into commodities and hard assets
→ Industrial demand surges (solar, EVs, infrastructure, defense)
→ China keeps pumping metals and reducing Western exposure
→ Western banks are trapped in shorts they can’t exit

This is a macro imbalance finally breaking.

If silver or gold starts moving even slightly - margin calls hit.

And they've just started.

In a market this tight, a short squeeze doesn’t mean “prices go up.”
It means prices RESET.

Gold reprices.
Silver reprices.
Commodities reprice.

Fiat currencies?
Infinite supply.

Metals in the ground?
Limited supply.

And while central banks globally race to dilute their currencies into oblivion, China is positioning itself for the aftermath.

With assets, metals, and leverage over global trade partners.

This isn’t a theory.
This is how empires transition.

A global market shock is coming, and almost no one is ready for it.

I’ve called nearly every major market top, including the October flash crash.

Follow, turn notifications on.

I’ll post the warning before it hits the news.

[The Kobeissi Letter @KobeissiLetter]

BREAKING: The 10-Year Note Yield officially drops below 4.00% for the first time since April 4th.

Markets are now fully pricing-in three 25 bps interest rate cuts by year-end.

[Chris @IntrinsicThink]

@KobeissiLetter Where is the logic? It is an attack on nuclear sites and not oil production sites. Even so, Iranian oil is under sanction.

Unless Iran decides to bomb the oil sites of its neighbors, then it is war. That may well be the end of Iranian regime. Would they be so dumb?

[Elon Musk @elonmusk]

False, this bill was never shown to me even once and was passed in the dead of night so fast that almost no one in Congress could even read it!

[Chris @IntrinsicThink]

@TrumpBarron__ No

[Wall St Engine @wallstengine]

SemiAnalysis published an analysis on DeepSeek, addressing recent claims about its cost and performance. $NVDA

The report states that the widely circulated $6M training cost for DeepSeek V3 is incorrect, as it only accounts for GPU pre-training expenses and excludes R&D, infrastructure, and other critical costs. According to their findings, DeepSeek’s total server CapEx is around $1.3B, with a significant portion allocated to maintaining and operating its GPU clusters.

The report also states that DeepSeek has access to roughly 50,000 Hopper GPUs, but clarifies that this does not mean 50,000 H100s, as some have suggested. Instead, it’s a mix of H800s, H100s, and the China-specific H20s, which NVIDIA has been producing in response to U.S. export restrictions. SemiAnalysis points out that DeepSeek operates its own datacenters and has a more streamlined structure compared to larger AI labs.

On performance, the report notes that R1 matches OpenAI’s o1 in reasoning tasks but is not the clear leader across all metrics. It also highlights that while DeepSeek has gained attention for its pricing and efficiency, Google’s Gemini Flash 2.0 is similarly capable and even cheaper when accessed through API.

A key innovation cited is Multi-Head Latent Attention (MLA), which significantly reduces inference costs by cutting KV cache usage by 93.3%. The report suggests that any improvements DeepSeek makes will likely be adopted by Western AI labs almost immediately.

SemiAnalysis also mentions that costs could fall another 5x by the end of the year, and that DeepSeek’s structure allows it to move quickly compared to larger, more bureaucratic AI labs. However, it notes that scaling up in the face of tightening U.S. export controls remains a challenge.

[Chris @IntrinsicThink]

@PeterSchiff Does this Trump understand what he is doing?

Americans will have to pay for it.

[Jukan @jukan05]

Morgan Stanley’s View on the DeepSeek Shock

1. Semiconductors

While DeepSeek’s success is unlikely to alter semiconductor investment plans significantly, there are several factors to consider.
Feedback from various industry sources consistently indicates that GPU deployment plans remain unaffected.
DeepSeek’s technology is impressive, but major CSPs have not neglected to invest in such technologies.

In fact, much of what we consider groundbreaking—such as training with FP8, multi-token prediction, MLA, custom PTX code, and GRPO reinforcement learning frameworks—either originated with DeepSeek v-2 and DeepSeek’s mathematical model six months ago or can be found in extensive AI research literature.
This underscores the importance of how these technologies are implemented, and DeepSeek provides efficient designs in every respect.

However, given the timing of announcements such as Stargate, Meta’s increased GPU demand forecast for 2025, Microsoft’s reiteration of its $80 billion annual CapEx guidance, and Reliance’s 3GW project in India, those connected to the model ecosystem were likely already aware of most of what DeepSeek was doing.

The long history of algorithmic improvements suggests we should not underestimate the incremental demand driven by cost reductions, advanced functionality, and continued scalability.
NVIDIA has stated that algorithmic efficiency has improved more than 1,000-fold over the past decade, surpassing the performance gains of single-chip inference. In this context, while DeepSeek’s proposal of a 10x reduction in training compute requirements may not significantly impact LTGR (long-term growth rates), the resulting cost savings could accelerate inference adoption and potentially increase demand for inference.

Export controls, however, remain a risk factor.
It is clear that restricting Chinese technology to H20-level performance has not halted China’s development of LLMs. The implications for government policy are unclear at this stage.
Lowering the performance threshold would greatly aid domestic silicon development in China. President Biden’s currently stalled AI restrictions, which aim to limit cluster sizes, would likely have a similar impact. These proposals may also compel other nations capable of supporting China’s AI development to obtain licenses.

Similarly, the performance gap between closed models and open-source ones continues to narrow.
We have highlighted three risks to the AI industry in our bearish outlook (AI remains a key driver, but smaller surprises are expected in 2025). One of these risks is that the number of companies in foundational model development could decline as it becomes increasingly difficult to compete with cheaper open-source options.

DeepSeek’s R1 exacerbates this risk by pressuring the largest spenders in AI to justify larger training runs while allowing others to leverage their work at much lower costs.

2. Internet: Lowering Barriers to Costs Drives Product Innovation and Adoption

DeepSeek’s architecture and pre-training improvements, which enhance cost efficiency, positively impact consumer internet companies seeking to develop new models and LLM-supported products.
The ROIC for GenAI CapEx is expected to increase, and incremental CapEx growth could slow as a result.

Larger companies’ ability to develop more innovative products will increase consumer utility, scalability, and adoption rates.
Given their large capital investments, user bases, and ability to extract and implement DeepSeek’s improvements into their own models, GOOGL, META, and AMZN are poised to benefit the most from these cost savings.

However, with more efficient architectures, smaller companies will also be able to provide GPU-supported products more broadly and at lower costs.
For example, AMZN’s AWS strategy focuses on commoditization at the model layer. AWS integrates third-party models and provides access through tools like Bedrock, enabling customers to build applications.

If DeepSeek’s contributions to democratizing model building (reducing required costs and compute) further commoditize models, AWS could benefit as an aggregator.

3. Software: Reducing AI Deployment Costs for Software Providers

Algorithmic efficiency gains at the model layer positively impact enterprise software.
More cost-efficient models are reducing the “GenAI deployment costs” for the broader software ecosystem, and the companies we cover are primarily building solutions around these models.

These efficiency improvements are not surprising to the software ecosystem. Most software companies believe that lower deployment costs lead to higher utilization (the Jevons paradox) and have already actively worked to lower these costs.

Microsoft recently focused on its “Phi” small language model (SML) strategy, with the Phi-4 14B model delivering benchmark results comparable to Llama-3.3 70B.

ServiceNow partnered with Nvidia to use custom domain-specific language models to execute inference more cost-effectively.

Snowflake trained Arctic 17B LLM with a $2 million training compute budget, achieving comparable performance to other enterprise benchmarks and best-in-class SQL performance.

Elastic developed the Elastic Learned Sparse EncodeR (ELSER) to lower the cost of semantic search for AI applications.

4. Energy

Regarding stocks exposed to the growth of AI power infrastructure in the U.S., significant capital expenditures are expected to continue.
Key considerations include:
1. An analysis of the U.S. data center pipeline indicates that most of the known pipeline is for AI inference and non-AI use cases rather than AI training.
2. The “Powering GenAI Models” analysis suggests that computing costs could drop by ~90% over the next six years. As AI adoption increases, the Jevons paradox could lead to rapidly growing demand for AI computing.
3. Discussions with companies suggest that substantial AI infrastructure spending is currently happening in the U.S. (Stargate being one of the most prominent projects we anticipate).
4. After the recent sell-off, many stocks exposed to U.S. AI infrastructure growth are still undervalued and not fully pricing in AI growth.

5. IT Hardware: The Focus of AI in IT Hardware and Apple’s Position

Last weekend’s DeepSeek news raised many unanswered questions (particularly regarding total compute costs and final training runs). Still, concerns about AI-related stocks’ long-term ripple effects are evident, especially for DELL (AI infrastructure) and STX (HDD), which may be the most impacted.

We believe $AAPL could emerge as a relative winner in this debate:

1. Apple’s AI ambitions are primarily focused on feature-specific, on-device small LLMs rather than large frontier models, meaning its AI investments are far less visible than those of its peers.

Consequently, Apple’s annual CapEx ($9.4 billion for FY24) is about 1/20th of the combined CapEx of U.S. Tier 1 hyperscalers. If the market overemphasizes CapEx ROI, Apple faces a much lower bar to generate attractive returns (i.e., less risk).

2. As DeepSeek has demonstrated, reduced memory requirements for inference make “Edge AI” much more feasible, aligning with Apple’s core GenAI ambitions.

3. In a world where consumer LLMs are commoditized, distribution platforms become key assets, and Apple owns arguably the most valuable consumer tech distribution platform in existence.

6. Embodied AI/Tesla: Advancements in GenAI Training Drive Embodied AI

In addition to the potential applications and acceleration of robotics training, we anticipate increasing attention on physical AI as growth in the digital AI narrative becomes less obvious and investors seek executable stories elsewhere.

In other words, as companies in the digital AI space become less reliant on double-digit returns, the early-stage opportunities in embodied AI—ranging from humanoids to eVTOL, AMRs, and AVs��are expected to appear increasingly attractive.

From a geopolitical perspective, autonomous vehicles currently operate in several cities at 25% of traditional taxi costs, supported by government policies encouraging innovation and supply chains for low-cost local production components. These developments once again highlight China’s achievements in embodied AI.

Faced with China’s advancements, the U.S.’s biggest geopolitical rival in all areas of AI, policymakers are expected to pay greater attention to fostering competitive progress among U.S. companies in this space.
$NVDA

[Thread Reader App @threadreaderapp]

@morganb Your thread is creating a buzz! #TopUnroll https://t.co/yonYt1lbfQ 🙏🏼@Beltranedlm for 🥇unroll

[Shay Boloor @StockSavvyShay]

Energy is the biggest bottleneck for AI data centers & meeting future demand will require innovative solutions to overcome infrastructure limits -- these are the key beneficiaries 🧐

• $TSLA -- Revolutionizing data center energy storage
• $FSLR -- Solar energy solutions for hyperscale data centers
• $CCJ -- Supplying uranium for next-generation nuclear data centers
• $CEG -- Delivering carbon-free energy for AI infrastructure
• $NNE -- Renewable energy driving AI data center growth
• $VST -- Powering reliable energy for data center operations
• $GEV -- Advancing energy efficiency in data centers
• $BWXT -- Picks and shovels for secure nuclear power infrastructure
• $SMR -- Modular nuclear power for modern data centers
• $TLN -- Renewable energy transition for sustainable data centers
• $VRT -- Cooling and power systems for AI-driven data centers
• $OKLO -- Compact nuclear solutions for decentralized data centers

[Shay Boloor @StockSavvyShay]

Trump officially unveils "Stargate," a $500B AI infrastructure initiative led by $MSFT OpenAI, SoftBank $ARM, $NVDA & $ORCL -- creating 100,000 jobs & advancing U.S. AI dominance under SoftBank chair Masayoshi Son with OpenAI managing operations.

[Dividend Talks on YouTube @DividendTalks]

4. Alphabet $GOOGL

Specialises in internet related services and products

• 30% ROIC

• 0 Net Debt to EBITDA

• Outperformed the S&P 500 over the last 10Y

• Cheapest valued stock in Mag7 currently

[Eric @EricFlaningam]

I wrote about AI data centers.

If hyperscaler capex is a good metric for AI infrastructure investments ($150B+ from $AMZN $META $MSFT $GOOGL over the last four quarters, up over 50% Y/Y), then we’re seeing one of the largest computing infrastructure buildouts in history.

Inevitably, with a buildout of this size, much of the supply chain is stretched thin. I think there are opportunities to address bottlenecks at each layer of the stack (energy, construction, compute infra, and compute services).

This infrastructure investment sets up the first half of the value creation equation. The second half comes with application value created on the back end.

Sharing more thoughts on the buildout below.

Note: this image doesn’t touch on every company exposed to the data center. There are financiers, real estate developers, construction firms, and a host of other companies contributing to this buildout. As Morgan Housel says, “I’m likely to agree with anyone who points out what I’ve missed.”

[Lin @Speculator_io]

$NVDA

Nvidia's Top Suppliers & Customers

Suppliers: Customers:
$FN - 35.1% $MSFT - 18.9%
Wistron - 17.8% $META - 10.1%
$TSMC - 10.1% $SMCI - 7.8%
$AMKR - 6.2% $GOOGL - 5.9%
SK Hynix - 5.3% $AMZN - 5.4%
$FLEX - 3.6% $DELL - 3.1%
ASE - 2.5% $TSLA - 2.1%

[Wall St Engine @wallstengine]

$ARM Holdings Q2 Earnings Highlights:

🔹 Revenue: $855M (Est. $808M) 🟢; UP +5% YoY
🔹 EPS: $0.30 (Est. $0.26) 🟢

Q3 Guidance
🔹 Revenue: $920M - $970M (Est. $945M)😐
🔹 Non-GAAP EPS: $0.32 - $0.36 (Est. $0.34)😐

Q2 Performance
🔹 Royalty Revenue: $514M; UP +23% YoY, driven by strong adoption of Armv9 architecture
🔹 License and Other Revenue: $330M; DOWN -15% YoY, reflecting timing of high-value agreements
🔹 Gross Profit: $820M; Gross Margin at 97.2%
🔹 Operating Income: $326M; Operating Margin at 39%, DOWN from 48% YoY due to increased investment in engineering

Segment and Strategic Highlights
🔹 CSS: Strong demand, with licensing growth from major partners in automotive and mobile, including new Dimensity 9400 processor based on Armv9
🔹 Automotive and IoT: Increased adoption of Armv9 for AI-driven applications across automotive, smartphones, and IoT sectors
🔹 Cloud and Data Center: Partnerships with Microsoft and Google on Arm-based Azure Cobalt 100 and Axion chips, enhancing energy efficiency and cost-effectiveness in data centers

Operational Metrics
🔹 Annualized Contract Value (ACV): $1.25B; UP +13% YoY, driven by strong license renewals and new agreements
🔹 RPO: $2.39B, maintaining stability with a slight decrease YoY
🔹 Free Cash Flow: $(65)M, impacted by deal-specific payment structures

CEO Rene Haas' Commentary
🔸 "Demand for high-performance Armv9 and CSS compute platforms continues to exceed expectations, fueled by AI proliferation from cloud to edge. Our strategic focus on energy-efficient compute solutions positions us well across sectors."

Strategic Developments
🔸 Arm and Meta Partnership: Optimization of Meta’s Llama 3.2 for Arm CPUs, leading to a 5x improvement in processing speed and enhanced energy efficiency
🔸 Ecosystem Growth: 20M+ software developers and new GitHub CoPilot integration, accelerating adoption and innovation on the Arm platform
🔸 Automotive Momentum: New CSS solutions for automotive, with projects underway with nearly one-third of the largest global automakers

[IvanaSPEAR @IvanaSpear]

$NVDA While everyone talks about Blackwell selling out, the real question remains: where will the next wave of demand come from?

Even investors like David Tepper "don't know how you know" about GPU demand beyond '25.

Here is something everyone is missing - let's dive in:

$AMD $MRVL $AVGO $TSM

[CN Wire @Sino_Market]

JUST IN
REUTERS SOURCES: CHINA'S LEADING LEGISLATIVE BODY WEIGHS APPROVAL OF NEW FISCAL PACKAGE EXCEEDING 10 TRILLION YUAN ON NOVEMBER 8

SOURCES: CHINA INTENDS TO APPROVE RAISING NEW 10 TRILLION YUAN DEBT THROUGH SPECIAL TREASURY AND LOCAL GOVERNMENT BONDS IN UPCOMING YEARS

FISCAL PLAN TO ALLOCATE 6 TRILLION YUAN FOR LOCAL GOVERNMENT DEBT AND UP TO 4 TRILLION YUAN FOR IDLE LAND AND PROPERTY ACQUISITION

CHINA COULD UNVEIL ENHANCED FISCAL MEASURES IF TRUMP SECURES U.S. PRESIDENCY.
https://t.co/kkBIzZRrmc
#CHINA $SHCOMP $SSEC $ASHR $KWEB $FXI $HXC $DRAG
https://t.co/8tAx8DYim9

[Chris @IntrinsicThink]

@gurgavin Sell !