Jack Lu

"Why it's more interesting to launch on Bittensor than start from scratch (Credit: @Axilo on X) Why TAO?  - Companies building on Bittensor are effectively subsidizing expensive compute by lifting on top of the network effects of Bittensor and its token, TAO as the primary incentive that everyone in the ecosystem competes for. - Most AI crypto protocols focus on deterministic outputs, rewarding only one type of result as the single source of truth. This results in one-dimensional incentive structures that are hard to scale and difficult to align with broader community participation. - Bittensor should be viewed not just as an AI protocol, but as an incentive protocol — a system where miners compete to produce subjective, high-value outputs across various subnets. - Because those outputs are subjective (e.g. language, vision, decision-making), the competition itself becomes dynamic, persistent, and difficult to game. - That creates a multi-layered economic game between miners, validators, and subnet owners with coordination, routing, and performance all incentivized natively - Many of these protocols are also hindered by high FDV funding rounds that front load investor returns and weaken the opportunity for meaningful, community driven growth. - Fairer launches like Bitcoin, Ethereum, and even Doge created stronger early ecosystems. Bittensor reintroduces that fairness with embedded incentive mechanisms that attract real contributors and incentives competition at every layer of the network. - As an investor excited about TAO, you're not betting on a single model or commodity. Instead on the collective intelligence / commodities of the network. - Every innovation from miners across every subnet is denominated in TAO, making it a meta-bet on decentralized AI productivity itself. - Instead of exposure to a single output, you're gaining exposure to the creation, distribution and competition for all outputs across the network. - As open-source AI models proliferate and centralized moats erode, Bittensor will become inevitable.. Offering open, incentive aligned coordination for AI development and deployment. - It's not just a bet on a network. it's a bet on the economic infrastructure layer of open-source AI. - What makes Bittensor especially compelling is how many existing crypto teams choose to launch subnets which proves that the economic design is working. I can see the same case for web2 companies coming to TAO to subsidize costs through miners. - It's simply more efficient to plug into TAO, align with existing miners and incentives, than to bootstrap a network from scratch. - Imagine if Midjourney, Stable Diffusion, or Hugging Face APIs were powered by miners competing to serve faster, cheaper, better results with the cost offset by TAO incentives. Whilst miners are hyper-competitive and have the incentives to do whatever it takes to outcompete their peers... That's a world I want to live in, and I believe that day will come."

12

4 Comments

Micron: Record 74% Gross Margin, 68% Operating Margin — but why did the stock fall after such a quarter? Micron delivered one of the strongest quarters across the semiconductor space, driven by a powerful combination of AI-driven demand, tight supply, and strong execution. Yet, despite record profitability and beat across key metrics, the stock declined post-earnings. Revenue reached $23.9B, well above expectations of $19.7B, supported by broad-based strength across segments. What stands out is not just growth, but the quality of earnings. Gross margin up to 75%, up more than 37 percentage points YoY, while operating margin reached 68%. Net margin followed at 58%. Free cash flow reached $6.9B, translating into a 23% margin, with EBITDA margin at 77%. This indicates not only strong pricing power but also disciplined cost control, with operating expenses remaining contained at roughly 6% of revenue combined for R&D and SG&A. AI workloads are significantly increasing memory intensity across all compute layers. Data centers are constrained not by compute, but by availability of DRAM and NAND, particularly for high-bandwidth memory. Micron is directly benefiting from this shift, with Cloud Memory and Data Center segments growing above 160–210% YoY, while even traditionally slower segments like automotive showed strong expansion. Product innovation reinforces this positioning. The ramp of HBM4 for next-generation AI platforms such as Nvidia Vera Rubin, alongside developments like 256GB LPCAMM2 modules enabling up to 2TB per CPU, highlights how memory is becoming a critical bottleneck in AI infrastructure. At the same time, Micron continues to expand manufacturing capacity globally, with new fabs in the U.S., Singapore, and India. Micron signed its first five-year Strategic Customer Agreement, signaling a shift toward longer-term demand visibility in what has historically been a cyclical business. Combined with a net cash position of $6.5B, Micron enters this phase with both financial strength and operational leverage. However, with such strong fundamentals, what is the market pricing in? Part of the answer likely lies in expectations. The guidance for the next quarter implies $33.5B revenue and 81% gross margin, suggesting continued acceleration. Capital intensity is rising sharply, with CapEx expected to exceed $25B in fiscal 2026 and increase further in 2027. Investors may be assessing how sustainable current margins are in the context of future supply expansion. Valuation remains relatively modest at 10x forward P/E and 7.6x EV/Sales, especially considering current profitability levels. The key debate shifts from near-term performance to cycle durability and long-term capital efficiency. Overall, the quarter reflects a structural shift in memory markets. The challenge now is not growth, but how efficiently that growth converts into sustained returns as capacity scales. #Micron #Semiconductors #AIInfrastructure #Investing #Earnings

38

2 Comments

📖 明快打字機: A Historian's Hunt for the Missing Link in Chinese Computing --A testament to human curiosity 🔗 [The New York Times](https://lnkd.in/gCF-3wJn) story by Veronique Greenwood In a fascinating global journey spanning 18 years, a dedicated historian unraveled the remarkable story of the MingKwai typewriter - a pivotal yet forgotten medium that bridged traditional Chinese language and modern computing. 🔍 Key Highlights: - 18-year research odyssey - Pioneering innovation before digital era - Technological breakthrough in language representation Further Exploration: - 🎙️ [NPR Article by Emily Feng](https://lnkd.in/gQbE4K7j) - 📝 [Made in China Journal by Yangyang Cheng](https://lnkd.in/gUQ_9fnR) #TechHistory #LinYutang

2

Nvidia, Foxconn reported to be in talks to deploy humanoid robots at Houston AI server making plan - https://lnkd.in/gYvJ-fWd These type of mid-market Humanoids Could Be Korea/Japan/Taiwan’s Moment: • China may be poised to dominate in consumer robotics but… • Korea, Taiwan, and Japan are ideally positioned to ship secure, reliable, mid-priced robotics and humanoids for more stringent applications • Cheaper than U.S. systems, safer than PRC supply chains: ideal for enterprise and government buyers Why this matters now: • Trusted supply chains + infosec baselines → KTJ ecosystems already meet industrial cybersecurity norms • Industrial-grade mechatronics + serviceability → legacy robotics and component expertise translate into uptime • Cost discipline without cutting safety → focus on modularity and repairability, not disposable hardware Recent catalysts: • Korea launched a national K-Humanoid Alliance to drive global leadership by 2030 • Taiwan’s robotics suppliers are emerging as preferred subsystem and sensor partners for U.S. and EU OEMs • Japanese firms are pivoting humanoid R&D into applied enterprise robotics: logistics, inspection, healthcare For manufacturing-tech investors: The humanoid shift isn’t about billion-dollar prototypes: it’s about scalable, serviceable, and secure humanoids built on KTJ’s manufacturing DNA. This is the mid-market inflection where enterprise robotics meets trusted industrial supply chains.

60

2 Comments

Every new NVIDIA chip cycle expands TAM/SAM far faster than it expands cost — and accelerated Big Beautiful Bill depreciation now allows companies to invest even quicker, reduce taxes, and reach the profitable end-state sooner. The major real systemic risk is a delay or underperformance in Rubin and Feynman, not rapid depreciation or fast iteration cycles Most investors are looking at the wrong AI risk. 👉 Depreciation is linear — AI demand creation is exponential. 👉 The faster the chip cycle, the better for those who can execute. What Rubin Creates Rubin isn’t just “faster Blackwell.” It enables entire industries that are not economically viable today: New Products Always-on AI agents, Autonomous multi-agent systems, AI-native systems that perform work, Real-time multimodal agents with memory, Continuous personal AI companions, Autonomous research and engineering tools New Services True RPA 2.0 automation, Autonomous ticketing & IT resolution, AI-native customer support, AI-driven knowledge management, Enterprise agent orchestration layers New Sectors AgentOps, AI governance & compliance, Agent marketplaces, Agent-native developer ecosystems Rubin is the bridge from the Copilot Economy (Blackwell) → Agent Economy. The Real Risk: Rubin Delays or Underperformance If depreciation is not the risk, what is? → Rubin arriving late → Rubin underdelivering → Rubin not enabling the agent economy on time Rubin is a platform inflection point. A delay would ripple across enterprise adoption, automation roadmaps, startup ecosystems, agent platform maturity, and hyperscaler consumption patterns. Early Internet Parallel: The Real Risk Then, and Now In 2000, one narrative was: “Mobile streaming is around the corner.” But the reality was: smartphones didn’t arrive until 2007 4G didn’t arrive until 2010–2012 cloud didn’t mature until 2014–2016 mobile video took a decade longer than investors expected The risk was not speedy investing. The risk was that the promised applications lagged the infrastructure by 8–12 years, and the enabled investments' main monetization channel in Y2K was something like clunky Yahoo ads. The Big Beautiful Bill The final piece of the puzzle is the new U.S. depreciation rules allowing maximum front-loading of AI infrastructure depreciation in the first year. This changes the financial profile of the AI build-out: 1️⃣ Companies can front-load depreciation → lower GAAP earnings in transition years → but free up cash flow → and reduce taxable income 2️⃣ Lower reported profits ≠ weaker economics It’s an accounting artifact, not a business deterioration. 3️⃣ Lower taxes + faster investment cycles = faster innovation More money stays inside the system, compounding. 4️⃣ Long-term profits will be larger and arrive sooner Because companies reach the agent economy earlier. 5️⃣ The market may misread the transition year(s) - 2026? Reported earnings will look weak when underlying economics improve the most.

8

2 Comments

A key discussion between Infleqtion’s CTO Pranav Gokhale and NVIDIA’s Sam Stanwyck on what shifted in 2024, and why hybrid quantum-classical computing is no longer merely theoretical. https://lnkd.in/eVRHxrZZ For BOKA, this matters greatly. Infleqtion (NYSE: INFQ) is helping to solve the integration challenge that has hindered quantum computing for decades: enabling GPUs and quantum processors to share workloads in real time. Quantum won’t replace classical compute, it will augment and increasingly enable it. QPUs won’t compete with GPUs; they will unlock new classes of problems that make GPU supercomputing more powerful. As Jensen Huang has said: “In the near future, every NVIDIA GPU scientific supercomputer will be hybrid, tightly coupled with quantum processors to expand what is possible with computing.” Logical qubits. Hybrid orchestration. Scalable deployment. The next computing cycle is taking shape with Infleqtion at the centre of it. #QuantumComputing #HybridComputing #HPC #DeepTech #AIInfrastructure #Infleqtion #NVIDIA

25

#NVIDIA agreed to pay $20 billion to license Groq’s AI chip technology and bring its core engineering team onboard, without fully acquiring the company. Groq will continue independently as #GroqCloud, but NVIDIA gains access to its LPU patents and leadership, including founder Jonathan Ross. The deal targets a key #market shift: inference now generates more revenue than training. While NVIDIA dominates training, it has lagged in low-latency inference. Groq’s #technology is built specifically for fast, real-time responses, which fills that gap. By structuring the deal as licensing and hiring, NVIDIA avoids antitrust issues, spends only about one quarter of its free #cash flow, and blocks competitors from accessing top inference talent. The move strengthens NVIDIA’s position as the end2end platform for #AI, from training models to running them in real time.

2