JPMorgan’s disclosure of a 5.25% stake in $SIVE carries profound structural and psychological implications that the broader market has entirely overlooked.
A critical detail in the disclosure is that JPMorgan accumulated a significant portion of this position through financial instruments and derivatives rather than the open market.
By locking up shares via swaps, they effectively removed a massive chunk of the tradable float quietly in the background. European and quantitative short sellers believed they were betting against retail fatigue, but they have unknowingly become the counterparty to a major U.S. institutional liquidity desk. The actual physical float is now far tighter than public short interest data suggests.
While critics point to the stock's massive year-to-date gains as a sign of a speculative bubble, they miss a regulatory reality. Many large U.S. long-only funds and institutions are legally restricted from purchasing companies below specific market capitalization thresholds. The recent price appreciation did not make the stock too expensive for institutions; it made it legally eligible for them. JPMorgan is not late to the rally; they are simply the first mega-fund permitted by regulatory mandates to establish a position.
This accumulation directly preceded $SIVE inclusion into major benchmarks like the OMX Stockholm Index and MSCI indices. Passive exchange-traded funds (ETFs) are now structurally forced to buy millions of shares to match these indices, regardless of valuation. By entering before these mandatory passive capital inflows, JPMorgan effectively frontran the structural buying pressure. Short sellers are trapped in a liquidity vacuum where they must cover into both a major U.S. investment bank and automated index buyers.
Beyond the commercial potential of the Co-Packaged Optics (CPO) supercycle for AI data centers, $SIVE holds a critical position within U.S. defense infrastructure through programs like the Pentagon's EW STAR initiative. JPMorgan’s entry signals that this technology is viewed through a geopolitical lens. U.S. institutional backing provides a defense mechanism against predatory short-selling on a company vital to Western infrastructure, validating the core retail thesis on a macroeconomic scale.
@asianinvestors I mean he is pushing the Fed not to, even thou they should, cause he knows his voters won’t like it. It’s funny considering the supply side inflation is entirely his fault rn
This is former CEO at Ayar Labs now at NVIDIA, commenting Ayar Labs CEO announcement about NVIDIA NVLink Fusion. Connecting well into the possible opportunity in this 3rd part that Goldman has yet to model.
I think my personal style of investing is a bit different, just some reflection:
It's inherently discretionary, based on stuff markets don't know yet. And a culmination of life experiences?
If you look at $AXTI, $RPI, $SIVE, $IQE and others.
Lot of it is guessing on unstructured relationships then seeing if it's right or not down the line.
$RPI is the perfect example:
1. Nobody really thought of Raspberry Pis for AI growth. Mainly people bought one or two just for class + education + hobbyist.
2. After OpenClaw, just noticed all my friends and people just buying Apple Mac Minis / RPIs for AI applications.
3. Found validation of that trend online with lot of people sharing video tutorials on AI orchestration with RPI.
4. AI was their ideal perfect growth vector, did some modeling, and thought it was compelling.
Earnings comes out and I was right.
Everyone in media was calling it a meme stock because there's nothing online that shows revenue growth from AI (was 14% forecasted revenue growth, turned out to be 58%, my projection was around 55%).
So it was a mix of guessing next industry trend (AI using lightweight hardware instead of GPU clusters), real life trends, then revenue forecasting off my guess.
For stuff like $AXTI:
1. Everyone called it a joke when I bought at ~$12. LLMs would hallucinate and say "hyperscalers/govs would have known about this by now and fixed this vulnerability with InP substrates"
2. Or would conflate very nuanced parts of InP substrate stack, where there's multiple different chokepoints in upstream processing.
3. So part of this was just discretionary based on what I've seen over InP substrate breakdowns, industry trends, etc.
4. Then also guessing the major supercycle was photonics (this was before everyone caught onto $LITE, and others). Or before you saw the $141B TAM projections from GS.
5. AXT owned 40% of InP supply chain, without them the supply chain just gets cripped).
6. All the "analysts" were forecasting steady InP substrate growth, few hundred million TAM, etc. or export controls.
7. Everyone kept trying to say $AXTI was overvalued based on TAM estimates. But if it's a few hundred million TAM you just think that's a joke and go into game theory over allocations.
8. Then I just had to guess, how much would this be worth if it were a NAND style bottleneck, what MC could it reach based on control, how much would hyperscalers price it as, etc.
A lot of the current research outputs from Goldman Sachs, or earnings reports from the Epiwafer companies, were confirmed after I published my piece on AXT. If you did research back then, lot of the same material /framing wouldn't have come up.
With stuff like $XFAB as you're seeing now, a lot of it is just pure guessing:
1. Not really any CPO materials, how much their MTP process makes in revenue, etc. Everyone online keeps saying they're not a photonics player.
2. But if you go through ASE docs or Gov websites, they all kinda cite XFAB as a major emerging player here.
3. $NVDA also evaluating them right now (maybe it's successful who knows).
4. No clear revenue around this area because their main silicon photonics process is still precommercial, but if you guess it's trying to create a EU supply chain to compete with $TSEM, once pre-commercial shifts to commercial, maybe similar but less volume contracts?
5. Then just seeing updates over the next few months to see if anything confirms this thesis guess.
_
I think a lot of information discovery still can be done with LLMs I'm seeing online. But it's also really hard to make a bunch of unstructured inferences based on unrelated material or even just trends you're seeing in real life.
So probably better to just do what's standard, eg. do valuation forecasting based on current numbers
Stuff like $AAOI, if they're projecting $471m/M h1 2027 and you see MC at $12B, probably undervalued might be a good idea to go long for next years.
Stuff like Samsung Electronics is easier, see what people are modeling for operating profits for 2027, 2028 then just seeing if it's undervalued or not at current levels.
Maybe something harder is $JBL. I haven't really seen any great volume numbers around 1.6T LRO, but you can just make a guess on how popular that might be then project how that might impact current MCs.
Or picking just good names everyone kinda agrees like $TSM, $INTC, $MRVL is also solid.
So a lot of things is just building up your life skills then applying that to markets. I don't think it's that can be taught with courses and stuff.
Of course, much of what I'm doing is just high conviction inference based on unconnected parts. Could always be wrong.
they'll probably shake you out, but they won't shake me out of $SIVE / $SIVEF
we understand the technology and how necessary for energy savings lasers can be for hyperscalers
watch what happens next