SMD_Contrarian

@jukan05 Do you really think memory trade is over?

“Sanjay has held a grudge against Apple since his days as a co-founder of SanDisk, and that resentment remains even now, in his role as Micron’s CEO.” It’s only natural that Sanjay bristles at the mere mention of Apple. What Apple did in 2023 was downright vicious. The days when it could procure NAND at cost and resell it to consumers at a 90% margin are over. Now that memory companies are fighting back, look at Apple acting like a spoiled child: “Hmph, fine—then I’ll just use Chinese memory!”

@jukan05 👍 agree

Apple is peddling the bullshit argument that iPhone prices directly affect inflation and that it needs to use Chinese-made memory to keep inflation low. Yet even though the fivefold increase in memory prices raised its costs by only about $50, Apple increased prices by $250. Cut your margins, Apple.

@AtlasShrug1 @jukan05 Why? History suggests Apple should have extremely high margin, but why not Apple should lower their margin and manufacturer should get higher margin?

@jukan05 Maybe the memory guys should cut their margins. Significantly.

@jukan05 Conclusion: Buy $NVDA when fintwit sell

Let me kindly explain to NVIDIA fanboys why Chinese open-source AI matters so much to NVIDIA. It’s because the proliferation of Chinese open-source models actually reinforces NVIDIA’s dominance. China is still training its models on NVIDIA GPUs and releasing them in forms optimized for NVIDIA hardware. Naturally, that makes NVIDIA GPUs the most advantageous platform for running inference on those models as well. After all, nobody is trying to run DeepSeek V4 or Kimi K3 inference on TPUs or Trainium. This dynamic helps protect NVIDIA’s moat in the inference market while also benefiting neocloud providers whose inference businesses are built on NVIDIA GPUs. It also gives NVIDIA a way to curb, at least to some extent, hyperscalers’ push toward their own custom ASICs. But what happens if China no longer has to depend on NVIDIA? NVIDIA’s dominance weakens, while the relative competitiveness of TPUs and Trainium grows stronger. It would also be bad news for the neocloud providers that NVIDIA regards as major strategic assets. So no, this isn’t simply about whether NVIDIA gets to sell chips to China or not, lol. If China achieves true compute independence, the ripple effects will be enormous.

@wangzhian8848 Bullish NV

梁文峰四小时交流会内容： 1，英伟达那套生态一年内就废了，华为的卡一年后完全能平替英伟达，贵一倍两倍都无所谓。 2，谁想着靠AI暴利谁就死，拿得少的人一定打败拿得多的人，OpenAI就是因为想独占才被围攻。 3，奇点不是什么爆炸，是个漫长的温水煮青蛙过程，而且顺序不能乱——先解决持续学习，再到自我迭代，最后才做机器人。 4，Ai技术路线是这样：语言模型 → 思维链(CoT) → Agent → 持续学习 → 智能奇点(自我迭代) → 具身智能，现在处于Agent走一半，持续学习是下一个必须跨过去的坎，跨过去才能进入AI自己迭代自己的奇点阶段。 5，克制是核心战略，主动放弃C端流量争夺战，不做超级App，短期商业化只是保底，全部资源主线聚焦AGI。 6，中美AI差距就一年，但国内只用了人家二十分之一的算力，人才根本不缺，差的就是卡。 7，公司一半的核心研究员都在标数据，这就是现阶段最重要的事，高端数据瓶颈不是钱，是时间。 8，最坏情况技术冻结了，光卖API也够撑起一家上市公司，商业化根本不用愁。 9，不加班、没KPI、没组织架构，全靠愿景驱动，因为做研究就得松弛，逼太紧反而没用。 10，国产芯片产能问题今年明年后年都卡，但五年后肯定解决了。 11，国内几十家做基模的公司太多了，最后就剩三四家，而且谁也别想有暴利。 12，视频生成、3D、世界模型这些东西跟智能上限没关系，纯粹是商业噱头，坚决不做。 13，员工期权拿够、团队稳住，AGI（通用人工智能）就一定能成，其他全是小事，最多晚个一年半载。 14，中国AI最终拼的是“系统性便宜”，就跟制造业一样，成本优势是结构性的。 15，持续学习要是先做出来了，通用智能就很简单，让AI自己帮自己研究就行。

@jukan05 @zephyr_z9 Very bullish NV

Wenfeng: - DeepSeek is working closely with Huawei and believes it can secure approximately 16,000 Huawei AI chips. - AI-powered code generation and advanced programming languages such as TileLang are expected to rapidly lower the barriers to entry created by the CUDA ecosystem. - Although NVIDIA GPUs were used to train V3, DeepSeek has significantly reduced its software dependence on NVIDIA’s ecosystem by using its own compiler and a TileLang-based environment. - If TileLang and DeepSeek’s proprietary compiler are ported to Huawei chips, the ecosystem issues surrounding Chinese chips could largely be resolved within about a year. The main remaining bottleneck would be production capacity. - Huawei’s 950 SuperNode could replace workloads currently handled by the GB200 and GB300, although roughly four Huawei cards are needed to match the performance of one NVIDIA card, and Huawei’s products remain about two years behind. - The chip gap between China and the United States is best understood as roughly a 4x gap in hardware efficiency and a two-year time lag—not primarily as an ecosystem gap. The end of CUDA’s moat is approaching. The ecosystem problem for Chinese chips could be solved within a year. The only real bottleneck left is production itself. Very bearish on $NVDA.

BofA’s Vivek Arya made an excellent point, which I’d like to share. Is open-source AI bearish for memory? Closed models amortize global demand across shared HBM pools concentrated in a handful of data centers, whereas open models create a new memory footprint with every deployment. If 10,000 companies self-host the same open model, the model weights must be replicated across 10,000 separate HBM pools, with each deployment also requiring its own KV cache. As 128K–1M token contexts become commonplace in 2026, the KV cache alone can exceed 40GB per active session. Low-cost Chinese APIs drive greater inference demand, broader enterprise self-hosting, and more memory sockets worldwide. In short, closed models concentrate memory demand, while open models multiply it.

@kevin__mak Isn’t DFTX looks like a better drug than CMPS?

$ATAI deal with Lily is pretty interesting, here's my opinion on the implications: First off, clearly bearish for Compass because, "Why didn't Lily buy $CMPS?". They would have evaluated the two and the transaction may be interpreted as favoring ATAI's platform, although Lilly's reasoning has not been publicly disclosed. Some investors may view this as a negative signal. Directionally negative/bearish doesn't mean absolutely bearish though. The likely reasons for Lily to pick ATAI instead: 1) They want to bet on the shorter acting drug 2) ATAI is probably cheaper to acquire than CMPS 3) They like the strategic profile ATAI's offerings over CMPS. (1) Makes sense as people worry about the 6-8 hour treatment regime for CMPS. I personally believe those worries are overblown. Plus, DFTX has a similar treatment regime and the market doesn’t seem too worried about that when trading it at a $6B valuation. (2) CMPS probably can't be acquired for less than $4-$5B. Not at this stage of the game. ATAI is going for $2.8B plus $1B CVR. So yes, it's cheaper, but this argument is pretty weak because Lily has a >1 Trillion dollar market cap so the price differences to them are largely rounding errors. (3) ATAI has a different drug mechanism and a few different avenues to commercialize. But ATAI is 2.5+ years behind Compass on approval (and meaningful risk of failure with the additional steps). --- On the bullish side, I think this increases the pressure on an acquiror, specifically JNJ to consider acting. They presumably want to defend their current Spravato brand/drug/market share. If Lily is taking ATAI, and if someone else takes Compass, it becomes a very challenging and competitive space in 2030 and beyond. A plausible defense here is that a potential acquisition of CMPS by JNJ is one of several possible outcomes and has ~3 years to really solidify their leadership in the market before Lily shows up. Plus it lets JNJ further monetize the REMS-clinic infrastructure that they spent so long building out for Spravato, instead of letting CMPS (or some CMPS acquiror) take advantage of it. Overall this development doesn't meaningfully change my thesis in either direction. My full writeup here: x.com/kevin__mak/sta… Disclosure: Long $CMPS, no position in $ATAI Creek Drive and its affiliated funds may hold positions in securities discussed. Views expressed are current opinions for informational educational only, subject to change and are not investment advice or a recommendation to buy or sell any security. Investing involves the risk of loss and past performance is not indicative of future performance. See the disclosures link above for more important information.

@Sunny_and_72 Do you hear anything about $CCOI related with the news? @RecurveCapital

Signs emerging

I‘m pumped that open-source models appear to perform almost on par with OpenAI and Anthropic. Not only is it nice to see human ingenuity win against massively bigger and better capitalized competitors (traditional David vs Goliath) - but it will also make sure AI is affordable and broadly accessible for the benefit of all of humanity.

Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out. Can't wait to hear what you build. Stay tuned! Token Plan ： international：qwencloud.com/pricing/token-… China：platform.qianwenai.com/pricing/token-…

@Biohazard3737 @GavinSBaker If time frame is 6-12q I think he’s implying optical, memory, data center owners, energy providers

@GavinSBaker Which ones? Or what sub-sectors?

Risk/reward seems attractive again. Lots of cheap stocks with durable competitive advantages that are going to crush numbers for the next 6-12 quarters. Time will tell!

Totally agree

@DeepSailCapital $MSFT

Assuming software revaluates over the next quarter, what are your most attractive software names?

@SemiAnalysis_ Time to be bullish when semianalysis get bearish, vice versa.

MASSIVE DELAY: Just 3 months after Jensen demoed Kyber NVL144 at GTC, it has faced major setbacks and has been delayed by more than 12 months, pushing it back to 2028. Below, we explain why Kyber has faced massive delays and why NVIDIA’s NVL72x2 back-to-back rack architecture was also cancelled, leaving Rubin Ultra with a limited scale-up domain. 👇️ 1/6🧵

@joedab12 @kaizen_cap Haha, If we ever disagree, let's just go with your opinion.

@matrix_trader_d @kaizen_cap if that were true the market cap would've been zero 4 years ago but it wasn't. silly take. cashflow matters.

If $META figures out AI, stock clearly not priced for it, and multiple higher. If it doesn’t, its DC footprint can be monetized, and EPS higher. Best torque out of the large caps. Seems assymetric here

@joedab12 @kaizen_cap Legacy business is given 0 valuation.

@matrix_trader_d @kaizen_cap oracles legacy biz sucks compared to metas

@LinQingV Is this good or bad for the industry?

It’s just beginning of dumping cheap token if we talking about two or three years time frame.