Chandan
4.5K posts

Chandan
@chandan1_
research crypto prev cofounder @fitcentive, investments in AI/web3 at @m31capital



according to silicon data, H100s are currently being rented out at $2.75/hr from neoclouds as a point of comparison, in 2023, a16z companies on the oxygen program were getting them at $1.72/hr

So, Jensen was right all along...


"Apple to Become World's #3 Notebook Vendor This Year"... Driven by MacBook Neo and Unified Memory Architecture Apple is projected to overtake Dell to claim the #3 spot in global notebook shipments this year, driven by the recently launched entry-level MacBook Neo and its Unified Memory Architecture (UMA). The global notebook market is expected to contract 8% this year, with Lenovo and Dell — both armed with strong memory procurement leverage — likely to outperform peers. On the 26th, market research firm Sigmaintell forecast that global notebook shipments this year will fall 8% to 181.1 million units, down from 196.7 million units last year. The decline reflects rising component prices, particularly memory chips, alongside broader market weakness. Vendor-level shipment forecasts for this year are as follows: Lenovo 43 million units, HP 39 million, Apple 28 million, Dell 22.5 million, Asus 16.5 million, Acer 10.9 million, and others 21.2 million. Apple is the sole major vendor expected to post growth, with shipments rising 22% from 23 million units last year to 28 million this year. Every other major vendor is projected to decline year-over-year. Dell's shipments are expected to drop from 24.2 million to 22.5 million units, allowing Apple to leapfrog Dell into third place. Apple has rounded out its MacBook lineup this year with the launch of the entry-level MacBook Neo. The previous MacBook lineup consisted of the Pro and Air, both priced above the Neo. The MacBook Neo starts at 990,000 won. Sigmaintell also highlighted Apple's Unified Memory Architecture (UMA). UMA is a design unique to Apple Silicon in which the CPU and GPU share memory resources. According to Sigmaintell, Apple has standardized memory specifications across multiple product lines, improving flexibility in component sourcing. Apple's revenue model is another differentiating factor. Apple generates substantial services revenue from the App Store, iCloud, and similar offerings, meaning it does not need to fully pass through memory price increases into finished-product pricing. Apple uses hardware like the MacBook as an "entry point" into its ecosystem, with services revenue offsetting hardware-side cost pressure. Sigmaintell noted that Apple's revenue model differentiates it from x86-camp Windows notebook vendors that rely on Intel and AMD CPUs. The x86 camp is heavily dependent on hardware sales. While Apple's notebook shipment forecast trails Lenovo and others, per-unit profit and overall margins are expected to exceed those of peers. Within the x86 camp, fortunes are expected to diverge this year due to differences in memory procurement power. Lenovo and Dell, both with sizable server businesses, are expected to outperform, while Asus and Acer — with minimal server market share — face greater headwinds. Servers consume far more memory per unit than notebooks, so a larger server business translates into stronger memory procurement leverage. Securing memory in bulk at relatively lower prices helps ease cost pressure on the notebook side. All major x86 notebook vendors are expected to see shipment declines this year, but the magnitude varies. Lenovo, with its high server market share, is forecast to decline 6% (45.6 million → 43 million units), and Dell 7% (24.2 million → 22.5 million units) — both narrower than the overall market's expected 8% decline. In contrast, HP is projected to decline 11% (43.7 million → 39 million units), Asus 10% (18.4 million → 16.5 million units), and Acer 15% (12.9 million → 10.9 million units) — all steeper than the broader market. Acer's heavy reliance on low-end products is a particular drag. Combined shipments from vendors outside this top tier are expected to plunge 27%, from 28.9 million units last year to 21.2 million units this year. $AAPL


You think the AI bottleneck is HBM. You think it’s CoWoS. You think it’s GB200 cables or 800G optics or the Arizona power grid. You are looking at the wrong layer of the stack…. The REAL bottleneck is epoxy resin paste. Specifically, Liquid Compression Molding compound EME-G, a goopy, beige, photosensitive thermosetting resin that gets squeegeed onto HBM stacks before the mold press comes down and cures it. Without this paste, the silicon dies in an HBM stack delaminate, the TSVs crack, and your $40,000 GPU becomes an expensive paperweight. Sumitomo Bakelite (4203.T) makes roughly 90% of the world’s supply. The other 10% is split between Nagase ChemteX and Hitachi Chemical, both of whom buy precursor chemicals from Sumitomo Bakelite. The moat is vertical. The resin formulation is a trade secret developed over 38 years of iteration. It contains a specific ratio of silica filler to bisphenol-F epoxy with a coefficient of thermal expansion tuned to within 0.3 ppm/°C of silicon. Get the ratio wrong by 2% and the HBM stack warps during reflow. Samsung tried to qualify a domestic Korean alternative in 2022. They failed. SK hynix tried in 2023. They failed. Micron didn’t even try. Sumitomo Bakelite ships approximately $180M of EME-G annually at a gross margin of 74%. Each HBM stack consumes roughly $6 of resin. As Nvidia’s roadmap points to more and more HBM stacks the math is clear. Feynman GPU package has 8 HBM stacks. That’s $48 of Sumitomo Bakelite content in a $70,000 GPU. They are 0.07% of the BOM and 100% of the bottleneck. This is the most asymmetric pricing power in the entire AI supply chain and they are charging like it’s commodity epoxy because the company is run by Japanese chemical engineers who think 8% annual price increases are aggressive. In Q3 of 2026 the Nvidia Rubin Ultra ramp is scheduled and will trigger an EME-G shortage. Sumitomo Bakelite will have to raises prices 35%. The stock will get re-rated from “specialty chemicals” to “AI infrastructure.” Multiple expansion from 14× P/E to 38× P/E. It’s a three-bagger in 18 months and the stock is up >3% YTD.





DeepSeek is a major threat to United States AI dominance.



i couldnt image myself working with a local model at 15tps





