STROKE at the age of 34.
Runs 5k on weekends.
Normal blood pressure.
No diabetes. No smoking.
But ONE common stress relief habit quietly tore a major blood vessel and caused a massive brain attack.
A medical thread on what actually happened 👇
Hey everyone, I have received permission from my employer to publish this research.
It is a lengthy investment thesis on what I believe will be the next chapter of AI. After months of research, we are turning bullish on the hyperscalers and explain why we believe the market is underestimating where the economics of AI are ultimately heading.
The rest is covered in the X article below. I hope you enjoy reading it, and I look forward to hearing your thoughts and challenges.
DBS, UOB, OCBC — Singapore's big 3 banks. Same decade. Same macro. Same rate cycle.
One returned +271%. Another +106%. The third +99%.
Here's everything you need to know. [1/8]
I'm a cardiologist. For months I've been telling you that inflammation is the fire behind heart disease — not just cholesterol. That we've been treating the smoke while the fire kept burning.
Penn Medicine just built the fire extinguisher.
They took the most powerful immune technology in cancer medicine — CAR T-cell therapy — and flipped it. Instead of engineering killer cells to destroy tumors, they engineered regulatory T cells to suppress the chronic arterial inflammation that causes heart attacks.
Published in Circulation. The results stopped me cold.
In CAR T cancer therapy, doctors extract your T cells, genetically engineer them to recognize a specific target, and infuse them back into your body as precision-guided missiles. It has cured previously terminal blood cancers. The technology won its developers a Nobel Prize.
The Penn team asked a question no one had asked before: what if we aimed this at the arteries?
They engineered regulatory T cells — Tregs, the immune cells whose job is to calm inflammation rather than cause it — to specifically target oxidized LDL. OxLDL is the molecule that starts the entire atherosclerotic cascade. It infiltrates your artery wall, triggers macrophages to gorge on it and become foam cells, releases inflammatory cytokines, recruits more immune cells, and builds the plaque that eventually ruptures and causes a heart attack.
OxLDL is the match that lights the fire. These engineered CAR Tregs are designed to find that match and snuff it out — at the arterial wall itself.
The results in mouse models of atherosclerosis:
Blocked macrophage foam cell formation — the cellular process that builds plaque. Dramatically reduced arterial wall inflammation. Prevented over 70% of plaque buildup compared to untreated controls. And critically — preserved normal immune function everywhere else.
That last point is essential. Previous anti-inflammatory approaches to atherosclerosis failed because they suppressed the entire immune system — leaving patients vulnerable to infections and other complications. Colchicine works modestly. Canakinumab in the CANTOS trial reduced events but increased fatal infections. The immune system is a sledgehammer. You can't just turn it down globally.
CAR Tregs solve this by being targeted. They don't suppress your whole immune system. They patrol your arteries specifically, calming the inflammation at the exact site where it's causing damage — and leaving the rest of your immunity intact.
One infusion. Targeted. Precise. The cells do the work.
Lead author Robert Schwab of Penn Medicine put it directly: "If we can get the immune system to see OxLDL and provoke an anti-inflammatory response, it would reduce inflammation and essentially stop the pathogenesis in its tracks."
Senior author Avery Posey: "Our study shows for the first time how CAR T cell technology could be used to treat the underlying cause of the most common form of heart disease — the leading cause of death worldwide."
As a cardiologist who has spent twenty years treating the downstream consequences of arterial inflammation — the stents, the bypasses, the cardiac rehab, the second heart attacks — I need you to understand what this represents.
Every treatment I currently have manages the damage after the fire has burned. Statins lower the fuel supply. Blood pressure meds reduce the mechanical stress. Stents prop open arteries that have already narrowed. These save lives. I use them daily.
But none of them put out the fire itself.
CAR Tregs are engineered to extinguish the inflammation at its source — inside the artery wall — before the plaque builds, before the vessel narrows, before the rupture, before the heart attack.
This is the shift from managing disease to correcting the biological process that causes it. At the cellular level. With living medicine.
This was demonstrated in mice, not humans. The leap from mouse models to human cardiovascular trials is enormous and filled with failures. Manufacturing CAR T cells is currently expensive — roughly $400,000 per treatment in cancer. Scaling this for a disease that affects billions would require a manufacturing revolution. Long-term safety of engineered immune cells patrolling human arteries for years or decades is completely unknown. Human trials are likely years away.
But 70% plaque reduction. With preserved immune function. Targeting the exact inflammatory mechanism I've been writing about for months. Published in Circulation — the flagship journal of the American Heart Association.
The trajectory is unmistakable.
Gene editing to permanently lower cholesterol. Personalized mRNA vaccines to hunt cancer. GLP-1 drugs rewiring metabolism. Cellular reprogramming to reverse aging. And now — living immune cells engineered to extinguish the inflammation that causes the number one killer on earth.
Every one of these treats the root cause instead of managing the downstream damage. Every one of them was impossible a decade ago. Every one of them is in trials or approaching trials right now.
I've held dying hearts in my hands in the cath lab at 3 AM. Hearts that were destroyed by inflammation I could see but couldn't stop.
The day I can infuse a patient with cells engineered to stop that inflammation before it ever builds the plaque — that's the day cardiology changes forever.
We're not there yet. But the fire extinguisher just passed its first test.
파이낸셜 타임스에 따르면 애플이 트럼프 행정부에 CXMT로부터 메모리를 구매할 수 있도록 허용해 달라고 로비하고 있다고 하네
이게 뜻하는바가 은근 큼
그냥 애플이 메모리 비싸다고 땡깡 부리는게 아니고 진짜 구하기 힘들다는 뜻이니까
로비를 시작한 게 가격 인상 발표보다 한 달 이상 먼저인데
이말인즉슨 아무리 협상하고 구할려고 해도 쉽지 않았다는거지
그러니까 정치 리스크 감수하고서라도 싸구려 비트라도 긁어모으려는 거고
근데 사실상 미국정부가 허락해준다 해도
쉽지는 않음
CXMT는 글로벌 DRAM의 약 5% 수준이고 점유율이 올해 6%에서 8% 정도로 올라가는 중인데,
거의 다 중국 내수(샤오미·화웨이·오포)랑 정부 우선순위로 팔고있음
애플한테 떼줄 수출 잉여가 별로 없다는 말임
그리고 어차피
애플도 플래그십이 아니라 중국 내수용
저가 라인(저가 Mac, HomePod, Apple TV) 일부만 노리는 거니까 메모리3사 입장에서 별로 상관도 없기도 함
플래그십 iPhone에 CXMT가 들어가려면 TSMC 패키징 라인에 다이를 공급해야 하는 구조인데 인증 통과에 시간 걸리는 일이니까
결국 애플 입장에선 크게 도움되는 글로벌 플래그십이 아니라 중국 내수용 저가라인에라도 좀 써보자 이건데 가격때문이라기엔 CXMT께 그닥 싼거도 아니고 진짜 구하기가 힘들어서 인거 같음
(+메모리3사 에게 땡깡)
그럼에도 불구하고 만약 미국정부가 이를 허가해주고 애플이 CXMT 메모리를 쓴다면
메모리 3사에 한가지 찝찝할수 있는건 2027~28년에 CXMT 신규 상하이 Fab(기존 허페이의 2~3배 규모) 램프가 붙고 메모리 사이클이 꺾일 때,
애플 같은 서구 블루칩이 끊어준 "인증 레퍼런스"가 CXMT한테 엄청난 무기가 된다는 거임
이미 Dell·HP·ASUS·에이서도 CXMT 문을 두드리고 있고
그러니까 이번 시도가 찝찝한 건 당장의 물량 위협이 아니라, CXMT한테 서구 시장 진입 인증서를 끊어줄 수 있다는 선례 리스크임
미국정부 화이팅👍
BREAKING: Claude can now run Stock Market research like a top consulting firm (for free).
Here are 10 Claude prompts that replace $100K/year stock analysts (Save for later)
NVIDIA CEO Jensen Huang said the “ChatGPT moment for general robotics” is coming.
The biggest winners may not be the robot makers.
They may be the companies building what’s inside the robot.
Robotics ecosystem:
• AI Brains → $NVDA $QCOM
• Sensors & Perception → $OUST $CGNX $ALGM $VPG
• Edge AI Inference → $AMBA $CEVA $LSCC
• Motors & Motion → $NJDCY $RBC $RRX $AME
• Precision Joints → 6324.T 6481.T $ALNT
• Power Electronics → $NVTS $TXN $WOLF $RNECY $IFNNY $STM $MPWR $ON
• Energy & Rare Earths → $ENS $MP $USAR $LYSCF $UUUU
If you’re interested in Physical AI and robotics, I shared my top 5 robotics stocks and a deeper dive here:
https://t.co/2QLIbzo8Zi
The humanoid robot market is projected to reach $7.5 trillion by 2050 and the real money is not in the companies assembling the robots (Save this).
It is in the components that every single robot on earth will need, regardless of which assembler wins.
Here are the companies that benefit from each layer.
Harmonic Drive Systems (HSYDF / 6324.T) controls roughly 85% of the global strain wave gear market, the compact, zero backlash gearboxes that go inside every robot joint requiring precision movement.
A single humanoid robot can use 20 to 40 of these gears and there is essentially no substitute at scale.
That is one of the most durable monopolies in any hardware supply chain right now.
Nabtesco (NCTKY / 6268.T) holds roughly 60% of the cycloidal reducer market, the heavy-duty version of the same type of gearbox used in the legs and load-bearing joints.
Its operating profit rose 60% year over year as humanoid and industrial robot orders surged.
Schaeffler AG (SFHLF) is rapidly becoming one of the most important actuator suppliers in this space, signing strategic partnerships with multiple humanoid robot makers to supply both strain wave and planetary gear actuators.
Schaeffler manufactures all components in-house, which gives it both margin control and supply chain reliability that competitors cannot easily replicate.
Nidec (NJDCY / 6594.T) is the world's largest electric motor company and is building out a full integrated 6-axis humanoid motion solution combining the motor, gearbox, and controller into a single unit.
At $2.6 trillion yen in annual revenue, it has the scale to be a dominant supplier as humanoid volumes ramp into the millions.
Ambarella (AMBA) is the edge AI chip company that processes the visual data directly on the robot, without needing a cloud connection.
Its CV7 chip, launched at CES 2026 on a 4nm process, is designed specifically for the multi-sensor perception workloads that humanoid vision systems require and it runs at a fraction of the power of conventional solutions.
Yaskawa Electric (YASKY / 6506.T) recently acquired Tokyo Robotics and saw its operating profit rise roughly 70% in its most recent fiscal year as it pivots aggressively toward the humanoid supply chain.
It is one of the few companies globally that can supply precision motion components at industrial scale already today.
The pattern across all of these names is the same.
They are not betting on one robot company winning but rather sell to all of them, they have structural supply constraints working in their favor and the volume ramp from 2027 onward will flow directly through their order books.
Milk Road is tracking every layer of the humanoid robot supply chain, from the memory inside the brain to the gears inside the joints and the companies positioned to win before the mainstream catches on.
Come join Milk Road Pro and get our full analysis every day using the link below!
$AMKR — Amkor Technology
What it does: $MU builds the HBM memory. Amkor makes it usable. It packages, stacks, bonds, and tests HBM beside GPUs where AI performance is won or lost.
The world's best memory is worthless if it can't be packaged into working HBM fast enough. Amkor isn't just a supplier anymore it's a critical bottleneck co-designing next-generation AI packages years before launch.
2-year move: $14 → $93 (+560%)
$FORM — FormFactor
What it does: Every Micron memory die is tested before shipping. FormFactor's probe cards verify each HBM layer at the wafer level before it becomes part of a multi-layer stack.
One defective die can ruin an entire 8–12 layer HBM stack. As AI memory becomes more complex, testing becomes exponentially more valuable. No testing. No yield. No HBM shipments.
2-year move: ~$20 → $159 (+700%+)
$RMBS — Rambus
What it does: Rambus supplies the IP and controller technology that allows Micron's memory to communicate with AI processors. Nearly every major memory company licenses its technology.
HBM is useless without a controller directing data traffic. Rambus is effectively the toll booth every AI memory transaction passes through and collects royalties while doing it.
2-year move: ~$40 → $174 (+350–400%)
Most investors only watch who makes AI memory. The biggest winners are often the companies making sure it actually works.
Yes, these 3 are still relatively cheap compared to $MU
♻️RESHARE this post and write 1 comment, I'll DM you call options for these so you can just buy and hold!
Look at the chart they just go up with $MU.
Stock market leverage in Asia is beginning to crack:
Stock trade defaults reported by securities firms in Taiwan rose to $62 million so far in June, the highest monthly total since data began in 2019.
A stock trade default occurs when investors who bought shares on margin fail to pay for those purchases, or when sellers fail to deliver the shares required to settle the trade.
Trade defaults have soared +300% over the last 2 months and now exceed the 2021 record by ~20%.
This comes as margin purchases have surged +160% over the last 12 months, to $19 billion, close to the all-time high set just before the 2000 Dot-Com crash.
By comparison, margin debt has risen +94% in South Korea over the same period compared to +50% in the final 12 months of the Dot-Com bubble.
Market leverage is out of control.
$WDC — HDD shortage, AI data hoarding, pricing power returning fast.
$VRT — Every AI rack needs liquid cooling; demand outpacing supply.
$SNDK — NAND squeeze just starting; same supply story as MU.
$BE — Fuel cells skip grid delays; hyperscalers need power NOW.
$INTC — Foundry capacity crunch; government-backed comeback story re-rating.
These are the big bottlenecks of AI right now: energy, memory, chip making, and cooling every one of these has $MU's exact set-up.
Follow me @itsmichaelluu for my exact entries I will add for earnings. I'll share it next week!
Most heart attacks happens between 6-8 AM.
A cardiologist with 28 years of clinical practice reveals why the first 10 minutes after you wake up are the most dangerous:
I'm a cardiologist. After 40, stop guessing about your health. These numbers tell you whether you're building a long, vibrant life — or quietly declining without knowing it.
I run these on myself. I run them on every patient I care about. Most are cheap bloodwork. All are available now. And together, they paint a picture no standard annual physical will ever give you. Print this. Bring it to your next appointment. Your 60-year-old self will thank you.
ꟷꟷꟷ
𝗙𝗮𝘀𝘁𝗶𝗻𝗴 𝗜𝗻𝘀𝘂𝗹𝗶𝗻
Target: below 5 μIU/mL. Ideal: 3-4.
This is the 10-year warning bell your standard panel completely misses. Your glucose and A1c can look "normal" for a decade while your pancreas is working overtime to keep them there. Fasting insulin catches insulin resistance 5-10 years before your A1c moves. By the time A1c rises, the damage is already extensive.
𝗛𝗢𝗠𝗔-𝗜𝗥
Target: below 1.0.
Calculated from fasting insulin and fasting glucose. The single best measure of insulin sensitivity. Above 1.0 and your metabolism is already under strain. Above 2.5 and you're insulin resistant — even if every other number looks fine.
𝗛𝗯𝗔𝟭𝗰
Target: below 5.4%.
Not below 5.7% — that's the threshold where medicine calls you "prediabetic." By then you've been metabolically compromised for years. Optimal is below 5.4%. Blood sugar mastery is longevity mastery.
𝗧𝗿𝗶𝗴𝗹𝘆𝗰𝗲𝗿𝗶𝗱𝗲 : 𝗛𝗗𝗟 𝗥𝗮𝘁𝗶𝗼
Target: below 2. Ideal: below 1.
Your metabolic health crystal ball. This ratio predicts insulin resistance, cardiovascular risk, and metabolic syndrome better than any single lipid number alone. A ratio above 3.5 is a red flag regardless of what your total cholesterol says.
𝗔𝗽𝗼𝗕
Target: below 80 mg/dL for moderate risk. Below 60 for high risk.
I've written about this extensively. ApoB counts every atherogenic particle hitting your artery walls. A 2024 analysis found 54% of patients had dangerous levels that standard LDL testing completely missed. If you only know your LDL, you're driving with one eye closed.
𝗟𝗽(𝗮)
Test once in your lifetime.
100% genetic. 1 in 5 Americans are elevated. Triples heart attack risk independently of everything else on this list. Diet and exercise cannot lower it. The 2026 ACC/AHA guidelines now recommend everyone be tested. Most never have been.
𝗵𝘀-𝗖𝗥𝗣
Target: below 1.0 mg/L.
You can have perfect cholesterol and inflamed arteries silently preparing to rupture. hs-CRP measures the fire behind the plaque. The JUPITER trial proved that finding and treating inflammation saves lives — even when lipids look fine. If this number is elevated, your mouth, your gut, your metabolic health, and your visceral fat are the first places to investigate.
𝗩𝗶𝘁𝗮𝗺𝗶𝗻 𝗗
Target: 50-80 ng/mL.
Not the bare minimum of 30 your doctor accepts. Suboptimal vitamin D is linked to higher inflammation, weaker immunity, increased cardiovascular events, worse mood, and poorer outcomes across nearly every disease I treat. Supplement D3 with K2 — without K2, calcium deposits in your arteries instead of your bones.
𝗧𝗲𝘀𝘁𝗼𝘀𝘁𝗲𝗿𝗼𝗻𝗲 (𝗧𝗼𝘁𝗮𝗹 + 𝗙𝗿𝗲𝗲)
Men: optimal range 600-1000+ ng/dL total.
Declining testosterone is an independent predictor of cardiovascular death in men. It's tied to insulin resistance, arterial stiffness, visceral fat accumulation, and systemic inflammation. DHEA-S drops 10-20% every decade after 30. Tracking these isn't about vanity — it's evaluating your body's systemic resilience.
𝗕𝗹𝗼𝗼𝗱 𝗣𝗿𝗲𝘀𝘀𝘂𝗿𝗲
Target: below 120/80. Aim closer to 110/70.
Every point above optimal is cumulative arterial damage. Buy a home cuff. Measure morning and evening, seated quietly for five minutes, arm at heart level. White-coat readings in the office miss what's really happening. The smartest $40 investment in cardiac self-care.
𝗩𝗢𝟮 𝗠𝗮𝘅
Men over 40: above 40 mL/kg/min. Women over 40: above 35.
Cardiorespiratory fitness is the single strongest predictor of all-cause mortality — stronger than smoking, diabetes, or heart disease as individual risk factors. A landmark study in JAMA found that extreme fitness was associated with the lowest mortality with no upper limit of benefit. You can estimate VO2 max with a timed mile, a rower test, or a wearable. Get faster every year.
𝗡𝘂𝗺𝗯𝗲𝗿 𝗼𝗳 𝗠𝗲𝗱𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀
Target: as few as possible.
Every medication you're on should be earning its place. I just wrote about five commonly prescribed drugs that do more harm than good with long-term use. Bring your full medication list to every appointment. Ask: "Do I still need this?" Deprescribing is one of the most powerful and underused tools in medicine.
ꟷꟷꟷ
Thirteen numbers. Most available through cheap bloodwork and simple tests. Get them once or twice a year. Here's what I want you to understand: these numbers don't just tell you where you are. They tell you where you're heading. A fasting insulin of 8 today becomes diabetes in five years. An ApoB of 120 today becomes a heart attack in ten. An hs-CRP of 3 today means your arteries are inflamed right now — regardless of how healthy you feel. The standard annual physical checks a fraction of these. It was designed to find disease that's already there. This panel finds the disease that's coming — years before it arrives.
What gets measured gets improved. Optimize with the foundation I write about every week on this platform:
Zone 2 cardio plus resistance training 3-4 times per week. High-protein whole-food nutrition. Sleep 7-9 hours — non-negotiable. Morning sunlight. Stress management. And the targeted supplements I've covered in detail — creatine, magnesium, CoQ10, D3+K2, glycine, omega-3, psyllium husk.
The breakthroughs coming in the next decade — gene editing for cholesterol, cellular reprogramming, senolytics that clear senescent "zombie" cells driving inflammation and aging, GLP-1 drugs rewriting metabolic medicine — will be most powerful for people who've already built the metabolic foundation to receive them.
The future of medicine is personalized. But it starts with knowing your numbers today. Print this list. Book the bloodwork. Own the data. Prevention isn't passive. It's the most aggressive thing you can do for the decades ahead.
Micron reports on June 24.
Past the headline revenue and the guidance,
"4 things" are worth a closer look.
1. First, where commodity DRAM prices go. HBM has been eating into wafers and process capacity, and commodity DRAM prices climbed with it. But the two reprice on different clocks. Commodity resets every quarter. HBM, once a year. So when the cycle turns, commodity DRAM falls first, quarter by quarter, while HBM stays locked to its annual contract and holds, then drops all at once at the next negotiation. Everyone's eyes are on HBM revenue. The cycle's direction shows up earlier in the commodity price fewer people are watching.
2. Second, whether the phrase "supply allocation" is still there. That phrase stands for a seller's market. If it disappears, the shortage is starting to ease. If multi-year contract customers grow instead, the seller's advantage is hardening.
3. Third, watch HBM's margin, not its revenue. As the base die moves to TSMC's logic foundry and packaging and yield costs pile on, rising revenue doesn't fall straight to profit. Whether they capture the price increase as actual margin, or not. What that does to a memory maker's stock goes without saying.
4. Fourth, how much further they tie NAND to AI demand. Micron has already named data-center SSDs and KV cache as AI growth. The question this time is whether they hold and deepen that, or lift NAND into a growth pillar of its own. That would confirm NAND has entered as a new supply bottleneck.
8 BLOOD TESTS THAT REVEAL HOW HEALTHY YOUR CELLS ARE
Most blood panels check if you’re sick.
They don’t check how well your cells are actually functioning.
Here are 8 tests that show you what’s happening at the cellular level, that your doctor probably isn’t ordering: 🧵