What I Read This Week…
xAI acquires X, Nvidia is developing light-based networking switches to connect millions of GPUs, the U.S. electrical grid is deteriorating, and cells use electricity to induce cellular processes
Watch our Interview with Treasury Secretary Scott Bessent
Watch our Interview with Commerce Secretary Howard Lutnick
Read our Short Dive: The Trump Administration’s Fiscal Strategy
Caught My Eye…
xAI has acquired X in an all-stock transaction valuing xAI at $80 billion and X at $33 billion under a new holding company that combines both entities. The synergy of this deal is that X's shareholders get a stake in xAI's increased valuation, and xAI gets full access to X's data and user network to train Grok. xAI raised $6 billion in late 2024 valuing xAI at $45 billion, and another $10 billion by early 2025 at a $75 billion valuation. This increase in valuation means that X investors now own shares in a rapidly appreciating AI company rather than just owning shares of a social media platform, fulfilling Elon's promise that X investors would receive 25% ownership in xAI. The advantage for xAI is that it gains full access to X's 10+ terabytes of real-time data for training, giving it a path for Grok to become the most current, up-to-date, and context-aware language model. xAI gains a competitive advantage at a time when other model makers face challenges: OpenAI and Google are dealing with copyright lawsuits related to training data, while Meta is challenged to improve pre-training for their Llama models.
Nvidia is developing new networking switches designed to connect millions of GPUs across data centers. Networking switches are devices that direct data flow between computers, GPUs, and storage units within and across data centers. These switches use a technology called silicon photonics, which transmits data as pulses of light instead of electrical signals. Silicon photonics chips convert electrical signals from GPUs into optical signals, sending data over fiber-optic cables that increase transmission speed, lower energy usage, and reduce signal loss compared to traditional copper cables. Nvidia integrates these optical components inside the switch package rather than using separate modules, further improving performance. The combination of light-based transmission and this integrated design allows Nvidia’s switches to achieve data transmission speeds of up to 1.6 terabits per second per port (around 2x the performance of traditional switches), breaking through the conventional barriers of slower data transmission and increased latency when scaling beyond hundreds of thousands of GPUs. Nvidia's photonic switches are planned for initial release in late 2025.
The U.S. electrical grid faces a dual crisis of deteriorating infrastructure and a shrinking skilled workforce. The grid's infrastructure is aging rapidly, with 70% of transmission lines over 25 years old and power transformers averaging more than 40 years in service – an age when failure becomes more common. This aging system needs extensive maintenance and replacement to maintain reliability, yet the technical workforce needed for this work is diminishing at the same time. About 25% of utility workers became eligible for retirement between 2017-2022, creating an experience gap where 56% of utility workers now have less than a decade of experience. The consequences are becoming apparent – in 2021, the average U.S. customer experienced 7.5 hours of electric service interruption, while the North American Electric Reliability Corporation warns that 19 states could face rolling blackouts during normal peak conditions within five years if these issues aren't addressed. The American Society of Civil Engineers estimates a $578 billion investment shortfall by 2033 to meet basic reliability and capacity needs.
Recent research indicates that cells use bioelectrical signals to induce and regulate various cellular processes. How does this work? Every cell in the body maintains a voltage across its membrane by controlling the flow of ions like potassium, sodium, and calcium. Healthy, fully developed cells usually have a more negative internal charge compared to stem cells or cancer cells, which tend to be more electrically neutral or "depolarized." The differences in voltage between cells influence what the cell does next, acting as the "software" of the body, directing the genetic "hardware." For example, changes in voltage can trigger waves of calcium to flow into the cell, which in turn flip certain genetic switches – telling the cell to grow, divide, specialize, or even stay dormant. In cancer research on frogs, scientists have uncovered that restoring the normal electrical charge of a cell can stop cancer cells from growing and cause tumors to shrink. These findings suggest that cancer may be, in part, a disease of miscommunication, where the electrical signals that normally maintain cellular order and function are disrupted.
Other Reading…
OpenAI’s Latest $40B Funding Round Hinges on For-Profit Restructuring by EOY (WSJ)
Anthropic and Databricks Teams Agree to Generate $100M of Revenue Together (WSJ)
AWS Pitching Trainium Chips as Alternative to H100s (TechRadar)
‘Don’t Study Coding Now,’ Says Replit CEO (The Financial Express)
Groq and PlayAI Partner to Bring Text-to-Speech Model for Real-Time Voice Apps (Groq)
Will Model Context Protocol Stick Around? (Steven Sinofsky)
Several Startups Close Rounds of $100M+ (Crunchbase)
Fidelity Plans to Launch Stablecoin in Digital Assets Push (Financial Times)
How India Shops Online 2025 (Bain & Company)
Inside arXiv—the Most Transformative Platform in All of Science (WIRED)
A Visual Guide to Critical Materials and Rare Earths (The Economist)
On X…
AI role : Efficiency in DMV
Driver License: my recent experience
North Carolina population jumped by millions in couple of years as NC is one of the fastest growing state in the country. There are same numbers of DMV locations as same as many years ago. How can same numbers of locations with limited staff can provide smooth service. Here is the question, is it worth to add more DMV locations with more staffs which takes time and hire train mor staffs.
Mostly more than 50% of clients return without job done because they didn’t take all required documents even though some people checked online required documents or took some wrong documents.
let’s say if the DMV app has features which check the documents by just taking documents pictures and do primary SCREENING and give you preliminary get to go green sign to visit DMV in person. So during preliminary SCREENING. APP will guide. This will make the real client with sufficient documentation and avoid time wasted. Then why do we need to add more locations and staff as APP AI can screen necessary documents rather than screened by DMV staff and asked to bring correct documents which saves time for both sides. There is always room for improvement rather than bragging service-demand etc.
I e-brake all other podcasts and double click on All In when it drops. GenX’r here who benefits a great deal from how you break things down and keep us in the loops.