I Speculated That Stanley Druckenmiller's Tesla Buy Was About FSD. Billionaire Investor Ron Baron Just Confirmed My Thesis.

Tesla reported 1.48 million active Full Self-Driving subscriptions in Q2, up 56% year over year, while more than 55% of new North American deliveries included FSD activation. Services revenue, which includes FSD subscriptions, rose 50% year over year to $4.6 billion, supporting the thesis that Tesla is building a higher-margin recurring software business. Investor Ron Baron cited FSD adoption as the key reason to buy Tesla, but the article notes TSLA remains expensive on conventional vehicle-based P/E and P/S metrics and that Robotaxi and Optimus remain unproven.

Analysis

The investable question is not whether FSD is software-like, but whether net subscription additions can become large enough to offset automotive gross-margin volatility. At a roughly $100/month price point, the cited installed base implies sub-$2B annualized run-rate before churn, promotions, revenue recognition, and regional mix—meaning the service narrative is strategically important but not yet sufficient to independently support TSLA's platform multiple. The next 1-3 month catalyst is therefore disclosure quality: paid versus trial activations, monthly churn, take rate by vehicle cohort, and incremental gross margin matter more than headline subscriber growth.

A higher FSD take rate can modestly improve Tesla's unit economics and reduce dependence on vehicle price cuts, but it also raises the regulatory/safety sensitivity of the equity. Any high-profile incident, adverse NHTSA action, or evidence that drivers treat activation as a short-lived trial would impair both revenue expectations and the autonomy-option multiple simultaneously. This is a materially asymmetric downside because TSLA still trades as though software scaling and eventual autonomous monetization reinforce each other.

The competitive read-through is mixed. BYD's scale and lower-cost hardware remain the primary pressure on Tesla's vehicle margin, while GM and F have less near-term software upside but also far lower expectations embedded in valuation. Tesla's proprietary fleet data is only monetizable if it demonstrably improves safety or functionality faster than competitors; subscriber growth without measurable capability progress is a marketing metric, not a moat.

Consensus may be over-crediting billionaire-holder commentary and conflating activated users with durable recurring revenue. The more constructive contrarian case is that even modest, verifiable software ARPU can stabilize consolidated margins before Robotaxi contributes; the bearish case is that FSD is being valued as both a subscription product and a future network, creating double counting. Treat the next earnings release as a KPI-validation event rather than a reason to chase sentiment.

Key Decisions for Investors

• Maintain a neutral-to-underweight TSLA stance into the next earnings report; do not add on commentary-driven strength. Upgrade only if paid FSD net adds, churn and service gross-margin disclosure support a credible >$2.5B annualized high-margin revenue trajectory, rather than activation growth alone.
• For defined-risk upside exposure, buy a 3-6 month TSLA call spread only after a pullback of 10-15% or following KPI confirmation; finance with an upside call strike near the prior valuation resistance. Thesis requires software revenue/margin acceleration, while a guidance cut or regulatory action is the exit trigger.
• Use a tactical pair of long GM or F / short TSLA in equal beta-adjusted dollars if TSLA rallies materially ahead of earnings without revised FSD monetization disclosures. The pair isolates Tesla's autonomy multiple premium from broad auto demand; cover if Tesla reports sustained paid conversion and improving automotive gross margin ex-credits.
• Set alerts for NHTSA investigations, material FSD price/promotional changes, and any split of trial versus paid users. These are the missing data points that determine whether the recurring-revenue thesis is investable; absent them, the article does not justify a standalone directional long.