$MU CFO: "We intend to increase our capital return. Over time, we expect to return 100% of our excess cash to shareholders."
Analysts: "So youโre saying 100% will go back to shareholders. I assume the vast majority of that is in buyback...โ
$MU CFO: โWe will hold what we believe is appropriate excess cash. Weโve always said that we intend to grow the dividend over time. You saw us do a 30% increase recently, but the principal capital return we have will be share repurchase."
Everyone is chasing the humanoid logos. I did what I did with AI Infra -- looked underneath the hood.
The money is not in the robot. It is in the ~40 parts that repeat inside every body, the ones almost nobody can build.
Here's the full list according to Goldman Sachs:
1. The brain // the intelligence:
$NVDA Nvidia -- Jetson + GR00T. The obvious one, lowest humanoid alpha.
$QCOM Qualcomm -- RB6, lower-power industrial.
Also: Tesla, $GOOGL, $META, iFlytek (https://t.co/KU8j5ePbkQ), Huawei (private)
2. Sensors & perception // how it sees and feels:
$OUST Ouster -- solid-state lidar pure-play.
$CGNX Cognex -- machine-vision incumbent.
$ALGM Allegro -- magnetic position sensors. The quiet winner, every joint needs them.
$VPG Vishay Precision -- strain gauges, force sensing.
$NOVT Novanta -- six-axis force-torque (ATI).
Also: Orbbec (https://t.co/EmUgsWEgkb), RoboSense (https://t.co/LhyDNpQfLV), $HSAI Hesai, Sunny Optical (https://t.co/bGc3smstll)
3. Edge AI inference // decisions on-device:
$AMBA Ambarella -- edge vision processors.
$LSCC Lattice -- low-power FPGAs for sensor fusion.
$CEVA Ceva -- DSP and inference IP. Pure licensing, pure obscurity.
4. Motors & motion // the muscles:
$NJDCY Nidec -- world's largest motor maker.
$AME Ametek -- precision instruments and motion.
$RRX Regal Rexnord -- motors and drives at scale.
$RBC RBC Bearings -- precision components.
Also: Moons' Electric (https://t.co/SU7teQIXWZ), Zhaowei (https://t.co/SfgqzHgLXa), Leadshine (https://t.co/QVkDM2gVw8), Veichi (https://t.co/gmK0xmcWww)
5. Joints & precision motion // human-like movement --
THE POTENTIAL BOTTLENECK:
6324.T Harmonic Drive -- strain wave gears. Every Optimus joint uses one.
6481.T THK -- planetary roller screws and linear bearings. Tokyo-listed, off the radar.
$ALNT Allient -- US-listed, integrated motion.
https://t.co/PItDtJ8PBJ Schaeffler -- roller screws.
Also: LeaderDrive (https://t.co/VL92TVG0Iw), Shuanghuan (https://t.co/8px4AQdQsO), Zhongda Leader (https://t.co/Znu2SoTgkf)
6. Dexterous hands // a humanoid in your palm:
$INVN... none US-listed. China leads here.
Zhaowei (https://t.co/SfgqzHgLXa) -- hand drive modules.
Inovance (https://t.co/kN8VYE6iDX) -- servo and motion control.
Also: PaXini (private, tactile)
7. Actuator assembly // putting it together:
Sanhua (https://t.co/4zM5co71yk) -- assembly at auto-supply-chain cost.
Tuopu (https://t.co/78h2i40K2N) -- built for cost and volume.
8. Power electronics // energy into controlled motion:
$NVTS Navitas -- GaN, the asymmetric small-cap.
$TXN Texas Instruments -- motor-control MCUs at scale.
$STM STMicro -- broad motor-control portfolio.
$ON onsemi -- power management and sensing.
$MPWR Monolithic Power -- high-efficiency DC-DC.
$IFNNY Infineon -- automotive-grade power.
$RNECY Renesas -- automotive and industrial motor control.
$WOLF Wolfspeed -- SiC pure-play, the distressed wildcard.
9. Energy & rare earth // the raw fuel:
$MP MP Materials -- the REE pure-play everyone knows.
$USAR USA Rare Earth -- magnet manufacturing, the new entrant.
$LYSCF Lynas -- largest non-China producer.
$UUUU Energy Fuels -- diversified REE and uranium.
$ENS EnerSys -- industrial batteries.
Also: CATL (https://t.co/havEzEZrGN). The magnet, not the cell, is the scarce part.
The one maker I'd own:
$CCXI Churchill Capital XI -- merging with Agility Robotics, trades as $AGLT on close. Only US-listed pure-play humanoid. Pre-deal SPAC, size it small.
The one maker I own: $CCXI / $AGLT
$CCXI Churchill Capital XI -- merging with Agility Robotics, trades as $AGLT on close. Only US-listed pure-play humanoid. Digit is real: 65k+ work hours, 300M+ in orders, backed by NVIDIA, Amazon, Foxconn.
The winners may not be the robot makers. They may be the companies inside the robot.
I made a deeper dive for free on my substack, go check it out:
https://t.co/euNn12SDVp
Now you know which floor to look at. โก
Gavin Baker @GavinSBaker: HBM DRAM will be 30-40% of all hyperscaler capex in 2027.
Baker manages Atreides - his 2025 Micron call is now 14x. This is his next call.
HBM requires stacking 12-16 DRAM dies in a single package. Only three companies can do it: Micron, SK Hynix, Samsung. No fourth supplier arrives in 2027. Hundreds of billions annually to three firms.
Micron's new supply chain agreements lock in floor pricing above prior cycle gross margin peaks.
If you still hold $MU at a commodity discount to $ASML and $LRCX, Baker says that discount is no longer earned.
Baker's full case at @theallinpod:
https://t.co/sH8c5iozAo
Source: All-In Podcast - https://t.co/31lFFUPJgJ
โNvidia upgrades its GPUs very aggressively, and HBM has to advance at the same pace,โ says former SK Hynix employee
I want to also make clear that the argument for the big 3 being replaced by cheaper memory is one of the most flawed arguments out there.
If Samsung, SK Hynix, and Micron themselves are constantly having to improve to keep up with higher complexity and demand; in example, to maintain Nvidia's compatibility for future products.
What makes you believe that any company can randomly pivot towards memory and outperform their products in such a short period of time?
Not even the fact that CXMT cannot physically put a dent on the supply/demand of memory, but even domestically speaking, no company can pivot towards memory at this point to give Micron any sort of turbulence
I think people are just having a hard time grasping seeing a legitimate change happening in tech after many years, and are over-complicating something which really is not complicated to begin with
The thesis for memory has not changed since last year, and that's the best part about it. If anything, one can argue that the true real world cycle for memory has just started as inference begins its boom
$MU $DRAM $NVDA
The memory supply-demand gap will keep widening through 2027. That is the real reason Apple is lobbying the White House to keep CXMT off the Entity List.
โStart with my latest industry checks: The pressure on Apple has shifted from soaring memory costs to a widening supply gap.
1. Of the memory capacity allocated to consumer electronics in 2026, an estimated 15โ20% is expected to shift to data centers in 2027, and that share could grow.
2. Due to tight memory (LPDDR) supply, Apple's actual pull-in volume of A20 chips in 2H26โ1Q27 could be 10โ20% below its original target (though part of that may reflect Appleโs own overbooking).
โCXMT states in its IPO prospectus that its capacity is far below domestic demand. Given the persistent global memory imbalance, even if Appleโs lobbying succeeds and it buys DRAM from CXMT, that would not materially lower costs or fill the supply gap. Still, with the imbalance widening, Apple has every reason to secure an additional source.
โThis also explains why Apple is being more proactive this time than it was when it evaluated YMTC in 2022. YMTC was mainly about lowering NAND costs; CXMT is about managing DRAM supply risk.
โTim Cook is one of the few tech leaders who can still navigate both Washington and Beijing, so this is better handled before he steps down as CEO. Even if the effort goes nowhere, the media coverage can still leave the market with the impression that Apple tried but was constrained by U.S. policy. That may help ease frustration over price hikes and longer delivery times.
Listen to what @GavinSBaker said on All-In at 1:04:06:
โDRAM is the most important bottleneck, and Elon is focusing the Terafab on memory because he believes memory is the most important bottleneck. Not lasers, not capacitors, not power semis, not NAND, not HDDs. DRAM.โ
Even Elon is prioritizing memory, specifically DRAM, as the key focus for the Terafab.
And yet, do you still think DRAM is just a commodity that can be replaced by Chinese products at any time?
This is Elon validating the strategic importance of DRAM.
@SouthernValue95 It literally doesn't matter? CXMT cannot satisfy China demand this decade. So there's 0nreason besides the small price arbitrage. It's like bannin Iranian oil. India and China buy it anyways, but price diff is less
๐จAPPLE AND MICRON GO HEAD TO HEAD IN AN ALL-OUT PUBLIC CLASH
Apple CEO Tim Cook went public blaming memory suppliers for the company's latest price hikes.
Micron's Chief Business Officer Sadana fired back, telling the WSJ that "certain customers" forced memory prices to rock-bottom levels during the last market downturn, preventing the Micron from investing as profits collapsed.
Without naming Apple directly, Micron executive says those aggressive pricing demands starved the industry of investment, forcing today's global memory shortage and the very price hikes now hitting consumers.
Apple spent a decade squeezing memory makers below cost. One of them survived, finally got pricing power, and Apple is calling it "gouging."
2 years ago $MU was selling DRAM below the cost of making it, gross margins underwater, in the kind of bust that has killed memory makers for 30 years. A field that once had dozens of players consolidated down to three survivors. Micron is here because it executed the downturns better than the ones that died, not because anyone was generous to it.
The buyers now screaming about 84.9% margins, Apple and the hyperscalers, are the exact players who spent that decade grinding memory to negative margins on every procurement cycle. Micron's own commercial chief said it out loud this week: blame the years of cheap procurement that made suppliers stop building capacity.
So this is not gouging and it is not collusion. It is the bill for a decade of underinvestment, and that underinvestment was a pricing failure. The actual service a speculator sells is marking a price close to intrinsic value, so capital flows to where it is needed. Analysts nailed Nvidia early; money poured in and supply got funded. They modelled memory as a commodity cycle right up until margins hit software levels. No accurate price meant no capital, no capital meant no capacity, and the squeeze now is the market collecting on its own mistake.
Memory stopped being a commodity the day LLMs took off. Compute happens in bursts and spreads across infrastructure; memory is the perpetual state the whole stack sits on. Chatbots to agents to real-time multimodal to robotics, every step up is built on more memory, not less.
If memory output 1000x's from here, the demand is already waiting to fill it. The supply glut everyone keeps pricing in is not coming back. This is the largest mispricing of an entire sector in the history of capitalism, and the gouging story is just the part where the market refuses to admit it was wrong.
$MU CEO ON ROBOTICS MEMORY DEMAND
โHumanoid robots carry 10 times the amount of memory as an average L2+ vehicle. We expect a sustained substantial multi-decade memory demand cycle to begin in the latter part of this decade.โ
โThe mix of L2+ and above vehicles is more than doubling this year to over 20%, and is expected to exceed 40% by 2030.โ
Supply Chain inside Ming-Chi Kuo says his latest checks show $GOOGL has no concrete plan to buy memory chips from CXMT.
The fact that rumors even surfaced around a Western hyperscaler considering Chinese DRAM shows how tight the memory market has become.