How to view FSDV12?

Tesla adjusted FSD to $99, which first impacts the Chinese market. The model of pre-installing hardware and charging for software is becoming difficult to sustain, severely damaging the long-term logic. On the first day of the Beijing Auto Show, a journalist asked a new automaker's CEO on my behalf: 'With Tesla at $99, does the logic of selling hardware to sell software still hold?' The CEO replied, 'That's a great question, but we can't answer it yet.' This incident, combined with what I saw at the Beijing Auto Show, left a strong impression on me. The impact comes from two aspects: First, the hyper-competitive environment—these cars are all incredibly strong. Beyond high-profile brands like Xiaomi and Li Auto, even second-tier brands offer impressive products for their price range. Second, the reality that no one is making money due to supply-demand mismatch. China's passenger vehicle industry operates at just over 60% capacity, and oversupply has shrunk the overall profit pool. Over the past five years, industry profit margins have declined from 6.3%, 6.2%, 6.1%, 5.7% to 5% in 2023, and this year it might drop below 5%. Profits are harder to come by. With FSD's price cut, the logic of 'cars don’t make money, software does' is broken in China. In North America, the most solvent market, it's only $99 per month—how much can they charge in China? Additionally, Chinese consumers prefer all-inclusive pricing, like a buffet. Introducing a paid item in a buffet setting won’t sell. Long-term, it might just be bundled into the car price. Yesterday, while discussing with Dr. Y, I noticed that even Chinese expats in the U.S. have low willingness to pay for software.

This is a classic case of being led into a pit by Tesla, much like the bike-sharing craze a few years ago—companies realized too late that the model didn’t suit them. Subscriptions aren’t generating revenue, and robotaxis? Setting aside technology, the labor markets in China and the U.S. are vastly different. U.S. labor is expensive, but whether China has a need for labor substitution is questionable. However, the impact of intelligence on the Chinese market is profound. I’ll share two observations: First, as mentioned earlier, revenue streams have turned into cost centers—intelligence hasn’t found a viable business model. Second, the core value of vehicles is shifting. Even if intelligence itself isn’t profitable, it’s increasingly becoming a key selling point. We’re seeing intelligence rapidly democratize—from ¥300,000 two years ago, to ¥250,000 last year, and now even ¥200,000 or lower offers excellent intelligent features. An extreme example is Baojun adopting DJI’s system at just ¥100,000. While an outlier, the trend is clear: without strong intelligence, cars won’t sell. This mirrors smartphones—hardware cycles are accelerating. The traditional three-year mid-cycle refresh and seven-year platform 换代 have at least halved in duration, while software 终身维护 becomes automakers' moat. New automakers exemplify this: hardware is homogenizing, while software creates differentiation and stickiness—even Huawei and Xiaomi are building ecosystems. Long-term, domestic competition will heavily favor in-house intelligent R&D; relying on suppliers slows iteration. The winners will likely be those with strong cash flow and distinctive, high-quality proprietary software—整车制造的门槛 isn’t high given China’s 供应链水平。

Returning to Tesla’s earnings this year: First, m2 wasn’t canceled—the most important Q1 news, securing future growth. Second, FSD has crossed the usability threshold. Reiterating an old point: assessing FSD’s lead over competitors shouldn’t use timeframes (e.g., '2-3 years'), as that implies catching up is inevitable. This is a generational gap. Industry PR versus engineering realities diverge sharply. With V12’s 端到端 approach, it’s now data-driven. Many have tested it in the U.S.—key takeaways: 1) Human-like behavior solves basic 体验 issues, encouraging 试用. Rule-based systems feel jarring (e.g., abrupt turns), alarming 普通用户. 2) MPI: our 讨论 concluded ~200 km between interventions is viable (weekly 接管 for me). Sub-50 km trips with zero 接管 suit commutes. 3) Most crucially, reaching L3 (no driver monitoring)—though 最难.

Tesla in 2024 feels like 2019—Musk all-in again. Many were shaken out in 2019, missing 2020’s rally. When Tesla debates peak，拐点 often follows.

FSD’s technical 拐点 is here, but pricing contradicts Musk’s 'getting pricier' claim. At 50% U.S. penetration (900k cars, +500k yearly), $500M revenue is underwhelming. As tech 落地，乐观 assumptions will adjust. Meanwhile, China—long deemed inaccessible—is pure 增量. Another realization: FSD’s value ties to usage 时长/里程. C 端’s issue is low user 时长. Solutions: 1) expand fleet (授权第三方 OEM). Key 拐点 is FSD becoming 黏性-driving 购车决策—a 体验 Tier1s can’t deliver. Imagine if only Huawei offered intelligence in China—would others buy its 套件？ 2) boost 单车里程 via ride-hailing. U.S.私家车 average 37 miles/day; taxis 150 (robotaxis 250+ with no downtime—7x 私家车).

假设 8.8 may unfold gradually: 1) recruit 车主 as safety 员 for 存量网约车 (FSD drives, Uber-like 收入). 2) deploy 无人车 in Waymo-approved zones. 3) partner with UBER—Elon’s vision isn’t 颠覆 UBER ($150B is small for Tesla), but leveraging its 触角. China-U.S. differences: EV 渗透率 (50% vs. single-digits) and 智能化渗透率 (Tesla’s few 100k vs. 150M annual U.S. sales—市场认知差距巨大). Uber enables rapid scaling. Long-term: remove steering wheels nationwide.

Of these 假设， C 端 is 2024’s brightest spot—I’m 乐观 on 北美 50%+ 留存 (cheap + 好用). Waymo-style 示范区无人车 may 落地 this year. OEM 授权 is long-term but inevitable. Tracking continues.

$Tesla(TSLA.US) $Li Auto(LI.US) $NVIDIA(NVDA.US)