RJC

Stop losses feel safe because they’re simple. But in a concentrated portfolio, they often force you to sell at the worst time, when volatility and fear are highest. If you’re going to be concentrated, the real discipline is thinking clearly not mechanical exits. open.substack.com/pub/rjccapital…

@aleabitoreddit This was filed 6 days ago

If you’re curious about $AXTI: It’s down 21% on the new potential dilution news. Board wanted to add 50M more shares (up to $2B worth to dilute) for shareholders vote in the 14th. 70m -> 120m shares. I say this about $IREN excessive $6B dilution and it’s the same with AXT price proposal that I hold. I would not “trust in management” to use it wisely and the fact it’s filed is a red flag. That being said: we’ll see what happens on the 14th. If it passes for 50m more share authorized dilution, I personally would not hold. But it fails, I expect a recovery.

Pete Hegseth clocking in for another hard days work

@danroberts0101 @NighthawkTradez is this bullish?

During routine testing of new equipment at Horizon, a suspected electrical fault in a diesel generator outside the data centers caused a fire contained to that single unit. The incident was contained within minutes, and no one was injured. With multiple spare generators on site, we do not expect any impact to GPU installation or customer handoff timelines. We appreciated the swift support from local response teams.

@jasonschips Lmao

Mediocre investors hate this one simple trick (h/t @RJCcapital)

@TheFinGodFather @retail_mourinho @NighthawkTradez @SJCapitalInvest I promoted it at $3 and had a 1% allocation lol can’t win them all

$NVDA $MRVL NVIDIA INVESTS $2B IN MARVELL TO EXPAND AI INFRASTRUCTURE AND SILICON PHOTONICS CAPABILITIES - WSJ Nvidia has invested $2 billion in Marvell as part of a strategic partnership focused on developing specialized AI chips and next-generation data center infrastructure. The collaboration centers on silicon photonics, a technology that uses light-based data transmission to improve speed and efficiency within AI systems. This approach is increasingly critical as data center bandwidth and power constraints become key bottlenecks. Under the agreement, Marvell will provide custom silicon solutions, while Nvidia integrates these into its broader AI ecosystem, including compute architecture and networking platforms. The partnership strengthens Nvidia’s vertical integration across AI infrastructure, extending beyond GPUs into custom silicon and interconnect technologies. It also positions Marvell to capture incremental demand tied to hyperscale and enterprise AI deployments. Strategically, the deal reflects a broader industry shift toward co-designed, application-specific hardware aimed at optimizing performance, power efficiency, and scalability in large-scale AI workloads. wsj.com/tech/nvidia-in…

$MU MICRON VERTICAL INTEGRATION STRATEGY EXPANDS WITH HBM-STYLE GDDR STACKING FOR AI MEMORY Micron is reportedly developing a new GDDR-based memory architecture that incorporates stacking techniques similar to high-bandwidth memory, aiming to improve throughput for AI accelerators while leveraging existing GDDR ecosystems. The initiative reflects a broader push toward vertical integration in advanced memory design, as Micron seeks to control more of the performance stack while reducing reliance on fully HBM-dependent solutions that remain supply constrained. The approach combines the cost and scalability advantages of GDDR with enhanced bandwidth density, positioning it as a potential middle-ground solution for AI workloads that require high performance without the full cost structure of HBM. This comes as demand for AI memory continues to accelerate across both training and inference, with customers increasingly seeking flexible and cost-efficient architectures amid ongoing supply tightness in high-end memory. If successful, hybrid GDDR stacking could expand Micron’s role across the AI hardware ecosystem, supporting broader adoption of advanced DRAM while reinforcing pricing and margin durability in performance-oriented memory segments.

$MSFT SECURES 900MW DATA CENTER CAPACITY FROM $ORCL AND OPENAI AFTER PROJECT EXIT - BLOOMBERG Microsoft has agreed to lease a large-scale data center project in Abilene, Texas, securing approximately 900 megawatts of capacity originally planned for Oracle and OpenAI. The shift follows the withdrawal of both companies from the site, allowing Microsoft to absorb a significant portion of pre-developed AI infrastructure. The facility is being developed by Crusoe and includes on-site power generation, with the first building expected to be operational by mid-2027. The project sits adjacent to the Stargate AI campus, a key hub supporting large-scale AI model training and inference workloads. The transaction highlights continued hyperscale demand for power-constrained data center capacity, particularly for AI workloads requiring high-density compute and dedicated energy supply. Leasing existing projects enables faster deployment relative to new builds. This development reinforces tightening supply across AI infrastructure, where power availability and permitting remain key constraints. Large pre-leasing agreements suggest sustained capital deployment from hyperscalers. Implications point to continued strength in data center development, energy infrastructure, and semiconductor demand, particularly across AI compute and memory supply chains. bloomberg.com/news/articles/…

$JPM $OWL $BX JPMORGAN LAUNCHES PRIVATE CREDIT FUND WITH ENHANCED LIQUIDITY TERMS VS PEERS - BLOOMBERG JPMorgan is planning a new interval fund focused on private credit that will allow investors to redeem up to 7.5% of assets quarterly, above the typical 5% cap seen across competing vehicles. The fund is also seeking regulatory approval to enable monthly repurchases of at least 2%, a notable shift from the standard quarterly liquidity structure that dominates the private credit market. This structure differentiates the product from peers, where stricter redemption gates have effectively limited investor withdrawals and highlighted structural illiquidity across the asset class. The widespread use of 5% redemption caps has contributed to rising investor concern, as funds have been forced to gate withdrawals, trapping capital and raising questions around liquidity mismatches in private credit vehicles. The portfolio will allocate at least 80% to credit, including a significant portion in private loans alongside publicly traded instruments, positioning it as a hybrid solution bridging private and liquid credit markets. bloomberg.com/news/articles/…

$META INCREASES AI INFRASTRUCTURE INVESTMENT WITH $10B EL PASO DATA CENTER EXPANSION - BLOOMBERG Meta Platforms is increasing investment in its El Paso, Texas data center to over $10 billion, up significantly from the initial $1.5 billion commitment, reflecting accelerating AI infrastructure demand. The facility is designed as a gigawatt-scale campus and is expected to come online in 2028, supporting more than 300 permanent roles and approximately 4,000 construction jobs at peak buildout. The expansion aligns with Meta’s broader capital expenditure outlook of up to $135 billion this year, with a growing share directed toward compute, networking, and data center capacity to support AI model training and inference workloads. Meta’s data center footprint now exceeds 30 sites globally, with newer builds increasingly targeting 1 gigawatt capacity, underscoring a shift toward hyperscale AI clusters with materially higher power and hardware density requirements. The step-up in spending highlights sustained demand visibility across the AI infrastructure stack, including semiconductors, power equipment, and cooling systems, as hyperscalers compete to scale compute capacity. bloomberg.com/news/articles/…

@BrunoHautenfau1 Yes, new articles coming soon for $10k portfolio update and more

@RJCcapital Are you doing the substack still?

If Iran closes both The Strait of Hormuz and The Bab al-Mandab Strait, 20-25 million barrels of oil will be affected per day That's 20-25% of the global oil supply

@Shabach17 You can find it in the substack

@extralovingcup $20 substack + discord

@RJCcapital How much is your discord?

@StonkShepherd You join through the substack. Once you subscribe on substack, you will find the link to the discord

@RJCcapital Perfect! But I don’t see your discord link. Was interested in Joining

10k challenge account is off to a hot start; up 10% this week

$MU MICRON $24B SINGAPORE FAB HIGHLIGHTS POWER INFRASTRUCTURE CONSTRAINTS IN ADVANCED MEMORY EXPANSION Micron’s planned $24B DRAM fabrication facility in Singapore is expected to require up to 500 electrical transformers, underscoring the growing power intensity of leading-edge semiconductor manufacturing. The scale of infrastructure reflects rising energy demands tied to advanced node production and AI-driven memory output. The facility is designed to support next-generation DRAM, including high-bandwidth memory, where process complexity and wafer throughput requirements significantly increase electricity consumption versus prior generations. Power delivery and grid readiness are emerging as critical gating factors for fab timelines. This development highlights a broader industry shift, where semiconductor capacity expansion is increasingly constrained not only by equipment and capital, but also by local energy infrastructure. Similar bottlenecks have begun to surface across major manufacturing hubs. For the memory sector, sustained AI demand is accelerating fab buildouts, but also raising execution risk tied to utilities and permitting. Power availability is becoming a key variable influencing supply growth cadence and regional allocation of future capacity.

@StonkShepherd Still going on. I made some updates yesterday. All of the alerts and positions are in the Discord

@RJCcapital @RJCcapital is the challenge still on? Or too late to join??

$GOOGL $MU TURBOQUANT DRIVES 6X MEMORY REDUCTION AND UP TO 8X GPU INFERENCE SPEED GAINS Google introduced TurboQuant, a compression algorithm targeting LLM key-value cache memory, delivering up to 6x memory reduction alongside as much as 8x improvement in inference speed while maintaining model accuracy. The approach optimizes how attention-state data is stored, addressing a core scaling bottleneck in large models. From a hardware utilization standpoint, the reduction in KV cache size directly lowers VRAM requirements per query, enabling more concurrent workloads per GPU. This translates into materially higher effective GPU utilization, with reported throughput gains of up to 8x driven by reduced memory bandwidth pressure and improved cache efficiency. The development is significant as inference increasingly becomes the dominant AI workload, where latency, throughput, and cost per query are critical. Software-level optimization allows existing GPU clusters to handle larger volumes without proportional hardware expansion. For memory semiconductors, implications are nuanced. Lower memory intensity per workload could ease near-term pressure on HBM and DRAM demand in inference-specific deployments. However, improved efficiency may accelerate adoption and total query volumes, supporting aggregate memory demand growth over time. Overall, TurboQuant reflects a shift toward algorithmic efficiency gains complementing hardware scaling, with direct implications for GPU utilization rates, AI cost curves, and memory consumption dynamics.

BREAKING: OPEN AI IS PLANNING TO DISCONTINUE SORA VIDEO PLATFORM, per WSJ

$CRCL COIN STABLECOIN REWARDS RISK DRIVES SHARP SELLOFF AS US REGULATORY LANGUAGE THREATENS USDC ECONOMICS - BLOOMBERG Circle shares fell as much as 22% in their steepest intraday drop on record, with the move driven by concerns that proposed US stablecoin legislation could weaken the economics of holding USDC. Coinbase also traded lower as investors reassessed stablecoin-linked revenue exposure. The key issue is draft Clarity Act language that could restrict exchanges from offering rewards on stablecoin balances. That would pressure one of USDC’s main distribution advantages, particularly as Coinbase currently offers certain customers 3.5% rewards on USDC holdings. For Circle, the concern is less about token stability and more about balance retention. If rewards are curtailed, stablecoins become less differentiated versus bank deposits, which could slow USDC growth and reduce incentives for larger holders to keep funds on platform. Competitive pressure is also building. Tether’s reported move toward a first full audit is fueling speculation around a broader US push, adding another headwind for Circle as regulation tightens and issuer competition increases. The broader takeaway is that stablecoin regulation remains a double-edged catalyst for crypto equities. Clearer rules may support adoption over time, but the market is now refocusing on how final legislation could shift economics across issuers, exchanges, and token holders. bloomberg.com/news/articles/…