Reward 500 Tiger-coinsOn the day RTX Spark was unveiled, the market had already cast its vote on this chain: foundry sole-winner $Taiwan Semiconductor Manufacturing(TSM)$ +4.1% closing at $436, memory leader $Micron Technology(MU)$ +6.6% closing at $1035; meanwhile $Qualcomm(QCOM)$, which had hoped to ride the Arm PC dividend with its own chips, got its rice bowl snatched away by Nvidia instead, down -8.8% in a single day closing at $229. One chip reshuffled the entire PC supply chain.
What exactly is RTX Spark?
Jointly developed by Nvidia and MediaTek, exclusively fabbed by TSMC on 3N: a 20-core CPU + 6144-core Blackwell GPU, the two sharing memory via NVLink, running Microsoft Windows for Arm. Launching this fall, with Dell, Lenovo, HP, Asus, Acer, and Microsoft all on board; initially positioned at the high end, aimed at AI developers, creators, and gamers.
Its significance isn't "yet another AI PC," but rather: the x86 duopoly that Intel and AMD have maintained on Windows for decades has, for the first time, been stabbed head-on by an Arm-camp super chip. Looking down this chain, the positions of who benefits and who gets hurt are crystal clear.
Foundry and equipment: the layer that wins lying down without even stepping onto the field
No matter whose chip wins out, as long as it's an advanced process, the foundry and upstream equipment make a sure profit. RTX Spark is exclusively fabbed by TSMC on 3N — $TSM$ is the unavoidable winner of this arms race, up +4.1% today; the long-term value of upstream equipment leader $ASML$ and test-and-packaging equipment maker Advantest also gets amplified accordingly. Meanwhile $QCOM$, also of the Arm camp and having hoped to ride the Windows on Arm dividend with its own chip, instead became the one whose rice bowl Nvidia snatched, plunging 8.8% in a single day — not everyone on this chain gets to eat meat.
Memory: shared memory is the most direct beneficiary
RTX Spark uses CPU and 6144-core GPU shared memory; for on-device large models to run smoothly, it places monstrous demands on memory bandwidth, capacity, and low power consumption. AI PC 2.0 will directly drive mainstream DRAM to leap from 16GB toward 32GB and 64GB as standard — this is a structural rise in both volume and price.
The big three $MU$, Samsung Electronics, and SK Hynix benefit most; high-speed large-capacity NAND is bullish for $WDC$ and $STX$, and flash supplier $SNDK$ is also on the core list.
Brand OEMs: high-end machines lift average order value and gross margin
On the launch list are $HP Inc(HPQ)$, $Microsoft(MSFT)$, Asus, $Dell Technologies Inc.(DELL)$, with Acer being the more purely PC-side player. Initially pushing the high end, directly lifting brands' ASP and gross margin.
$Apple(AAPL)$ guards its own M-series ecosystem, but the wave Nvidia is stirring up in Windows will force Apple to accelerate monetizing on-device AI.
PC-side AI transformation: which segment of the upstream/downstream are you most bullish on?
Do you think $MU$ can still be chased on this move, or has it already over-priced AI PC expectations?
Would you rather buy certainty, or bet on elasticity?
Leave your comments to win tiger coins~
For SG users only, welcome to open a CBA today and enjoy access to a trading limit of up to SGD 20,000 with unlimited trading on SG, HK, and US stocks, as well as ETFs.
🎉Cash Boost Account Now Supports 35,000+ Stocks & ETFs – Greater Flexibility Now
-
Find out more here.
Complete your first Cash Boost Account trade with a trade amount of ≥ SGD1000* to get SGD 688 stock vouchers*! The trade can be executed using any payment type available under the Cash Boost Account: Cash, CPF, SRS, or CDP.
Comments
I’m most bullish on the memory layer. AI workloads on-device need much higher capacity and bandwidth, and that shift looks structural rather than cyclical. That’s why I still like $Micron Technology(MU)$ , SK Hynix, and Samsung. Even after the rally, I don’t think AI PC demand is fully priced in if 32GB–64GB becomes mainstream.
Between certainty and elasticity, I lean toward certainty. TSMC remains my highest-conviction pick because it wins regardless of chip design outcomes. Nvidia may lead today, but no matter who dominates, advanced chips still need to be manufactured, and TSMC stays at the center of that value chain.
@TigerStars @Tiger_comments @Tiger_SG @TigerClub
These 2 companies own the un-bypassable intellectual property driving the next generation AI PCs. It does not matter which PC brand wins the retail marketing wars. Marvell & Broadcom act as the toll collectors on the internal highway systems of every AI PC manufactured.
Marvell completely dominates the design of high speed PCI Express Gen 6 & Gen 7 storage controllers. These controllers have fundamentally transformed how fast internal computer components can pass data to each other.
Broadcom is the master of co-packaged custom silicon. It holds a near monopoly on hyper advanced Physical Layer (PHY) & custom ASIC connectivity designs.
Broadcom simply collects high margin, recurring licensing taxes & component fees from chip developers before a computer is even sold.
@Tiger_SG @Tiger_comments @TigerStars
RTX Spark feels less about one chip and more about the whole AI supply chain shifting. Nvidia gets the attention, but companies like TSMC actually make the chips, so they seem like steady winners no matter who leads the AI PC race.
Micron also stands out to me. If AI laptops really need more memory (like 16GB → 32GB+), demand should naturally rise over time.
Qualcomm dropping shows how fast sentiment can change in this space.
As a beginner, I’d rather focus on “picks and shovels” than try to pick the winning AI device.
Curious if I’m oversimplifying this 🤔