Tesla (TSLA 7.21%) shares have bounced in recent weeks as investors refocus on the company's artificial intelligence (AI) ambitions alongside its core electric vehicle (EV) business. The EV maker and energy company is pushing to commercialize an autonomous ride-hailing network, dubbed Robotaxi, that it says will lean on the same vision system already shipping in its cars.
Against that backdrop, the stock's rebound has reopened the debate: Without LIDAR (Light Detection and Ranging), a sensor system that uses lasers to measure distances and create 3D maps of surroundings, can Tesla really make self-driving work at scale?
What Elon Musk said late last year and then again this spring goes straight to the heart of that question. It also explains why Tesla continues to double down on a vision-only approach despite others continuing to pursue more sensor-heavy stacks.
Image source: Getty Images.
The investment thesis hinges on autonomy
Investors should first anchor on the business today. In the second quarter of 2025, Tesla's operating income fell 42% year over year to about $0.9 billion, producing a 4.1% operating margin, as pricing pressure and mix weighed on automotive profitability. Tesla produced over 410,000 vehicles and delivered more than 384,000 vehicles in the quarter, while energy storage deployments hit 9.6 gigawatt-hours (GWh) -- a bright spot as the nascent but important segment scales.
The first quarter painted a similar picture of near-term pressure in the core auto business, with revenue down year over year and management emphasizing cost work and software progress as offsets. Shares, meanwhile, have rallied from summer lows and recently traded at around $395, putting Tesla's market capitalization at around the $1.3 trillion mark, as investors again assign significant optionality to autonomy and artificial intelligence (AI).
Those fundamentals matter because they frame why autonomy is so central to the long-term story. If Tesla can layer higher-margin software revenue (Full Self-Driving subscriptions, ride-hailing take rates) on top of a large fleet, the earnings profile looks very different from that of a pure automaker.
Why Tesla is staying vision-only -- and what that means for investors
On Tesla's fourth-quarter 2024 earnings call, CEO Elon Musk distilled the company's thesis for a vision approach to autonomy -- without LIDAR -- in plain language:
Obviously, humans drive without shooting lasers out of their eyes. ... [H]umans drive with eyes and a neural net. ... The digital equivalent of eyes and a brain are cameras and digital neural nets or AI. So, the entire road system was designed for passive optical neural nets.
He reiterated the same idea on the first-quarter 2025 call: The car, he said, is analogous to a human -- digital neural nets plus cameras versus biological neural nets plus eyes -- implying the same strengths and weaknesses.
The practical takeaway is clear: Tesla does not intend to add LIDAR as a prerequisite for wide deployment. A pure vision stack simplifies hardware, lowers bill-of-materials costs, scales with the installed base, and, if it works, expands margins through software leverage -- without the added cost of LIDAR systems. It also aligns with how the company trains its models: by harvesting billions of miles of real-world video from its fleet to improve neural-net-only perception and planning.
Of course, a vision-only system must prove sufficient across edge cases -- such as adverse weather, unusual road geometry, and unpredictable human behavior -- where redundancy from LIDAR and high-definition mapping can potentially help competing stacks. But there will be weaknesses to any system. Investors, therefore, will have to hope those weaknesses are minor and that the system is far safer than human driving.
Because regulatory approvals are not guaranteed, even limited driver-monitoring or operating-domain constraints could slow a robotaxi launch. And while Tesla emphasizes software, recent quarters show that automotive margin pressure can persist if pricing remains competitive and software (e.g., paid Full Self-Driving) adoption lags expectations.
For investors, the question is not whether LIDAR "wins" in a lab. The question is whether Tesla can achieve safe, scalable autonomy -- with acceptable regulatory guardrails -- using cameras and neural nets, and then monetize it at meaningful rates on a massive installed base. If it does, the long-term earnings power looks far more like a software and network business layered on top of manufacturing. If it does not, the growth stock's rich valuation already embeds optimism that could be hard to defend on vehicles alone.
Therefore, Musk's comment is worth hearing because it clarifies the bet that Tesla bulls are making. Tesla is pursuing the most capital-efficient autonomy path tied to how humans actually drive. That approach could translate into faster deployment and better unit economics if vision-only performance crosses the safety threshold that regulators and riders demand. But it also raises the bar on software progress and real-world validation data in the coming quarters.
In the quarters to come, Tesla investors should closely watch software take-rate trends, energy storage scaling, operating margins, and any concrete milestones on Robotaxi. The return profile here increasingly hinges on software and autonomy delivering -- not merely on selling more cars.
