News
AI dominates China’s elite doctors in cancer diagnosis competition
A custom-built AI designed to diagnose brain tumors and predict hematoma expansion dominated some of China’s best doctors in a competition last Saturday in Beijing. The AI, dubbed BioMind, ultimately scored 2:0 against its human competitors, comprised of 15 senior doctors from China’s premier hospitals.
BioMind was developed by a collaboration between a team from the Artificial Intelligence Research Center for Neurological Disorders at the Beijing Tiantan Hospital and researchers from the Capital Medical University. BioMind’s developers opted to feed the AI with data sets featuring tens of thousands of images depicting nervous-system-related diseases, which were retrieved from Tiantan Hospital’s archives stretching over the past decade.
Wang Yongjun, executive vice-president of Tiantan Hospital, stated that this training ultimately enabled the AI to become proficient in diagnosing neurological diseases such as meningioma and glioma with an accuracy rate of over 90%. According to Wang, such rates are comparable to the accuracy of a senior doctor, according to a report from state-owned Xinhua News.
During its the competition on Saturday, BioMind was able to correctly diagnose brain tumors with an accuracy rate of 87% out of a total of 225 cases. The AI was also able to complete its task in 15 minutes. In comparison, the team of 15 elite doctors was able to achieve an accuracy rate of 66% when diagnosing brain tumors, finishing the task in 30 minutes. Apart from this, BioMind was able to make correct predictions in 83% of brain hematoma expansion cases, while its human competition displayed a more conservative 63% accuracy.
Despite the AI’s strong performance against China’s elite doctors on Saturday, however, Cheng Jingliang, a professor of radiology at the First Affiliated Hospital of Zhengzhou University, stated that artificial intelligence systems for the medical field are still well into their infancy. According to Cheng, AI is already being used in hospitals to help doctors read images such as lung scans, but when it comes to giving full diagnoses to patients, artificial intelligence still lags far behind that of senior medical professionals.
In a statement to China Daily, Paul Parizel from the Antwerp University Hospital in Belgium, who served as a member of the jury during last Saturday’s AI vs. human doctors competition, believes that systems such as BioMind would prove to be incredibly valuable when integrated to existing medical practices.
“It will be like a GPS guiding a car. It will make proposals to a doctor and help the doctor diagnose. But it will be the doctor who ultimately decides, as there are a number of factors that a machine cannot take into consideration, such as a patient’s state of health and family situation,” he said.
The United States initially led the artificial intelligence race, but over the years, China has steadily gained ground in the AI industry. Thanks to a population that is compliant to the application of new technologies, as well as a government that actively pushes AI researchers to push further, China is on track to overtake the United States in the near future. Last January alone, the Chinese government announced plans to build a $2.1 billion technopark in Beijing that is expected to house companies actively involved in AI research and development. The United States does not have a comparable initiative to date. This was confirmed by Jack Clark of Elon Musk-backed OpenAI, who previously stated that the country lacks a central national strategy on artificial intelligence.
“It is confusing that we have this technology of such obvious power and merit and we are not hearing full-throated support, including financial support,” Clark said.
In early 2026, Tesla made a bold strategic pivot: it largely eliminated the option to purchase Full Self-Driving (FSD) software outright and shifted to a subscription-only model. The change, effective around mid-February, ended the one-time fee that had previously ranged as high as $15,000 and later dropped to $8,000. Instead, customers would access FSD (Supervised) for $99 per month in the U.S.
At the time, skeptics questioned whether locking customers into recurring payments would hurt adoption or alienate buyers who preferred ownership of the feature. Tesla bet that a lower barrier to entry, seamless integration at purchase, and the ability to cancel at any time would drive higher uptake.
The results from Q2 2026 speak for themselves: the decision has been a resounding success, delivering the largest quarterly growth in FSD subscriptions in the company’s history.
Tesla FSD subscriptions went up 56% in Q2 2026 to 1.48 million, an increase of 200,000 from Q1 2026.
Tesla added more FSD subscribers in Q2 than in any quarter in its history. pic.twitter.com/jTciTD2JqW
— Sawyer Merritt (@SawyerMerritt) July 22, 2026
According to Tesla’s Q2 shareholder update, active FSD subscriptions reached 1.48 million globally by the end of June 2026. That represents a 56 percent increase year-over-year and a 15.6 percent jump from the prior quarter. Tesla added roughly 200,000 new subscriptions in the period alone—the biggest single-quarter gain on record.
North America led the charge, with more than 55 percent of new vehicle deliveries including an FSD subscription at the time of purchase, a record attach rate for the region.
Tesla explicitly noted that “more customers [are] opting for subscription at the time of vehicle purchase,” crediting the model shift and prominent placement of the option in the ordering process. Subscriptions now contribute meaningfully to ancillary revenue, helping offset pressure elsewhere in the business.
The financial upside is substantial: At $99 per month, 1.48 million active subscriptions generate approximately $146.5 million in monthly recurring revenue. Over a full year, that equates to roughly $1.76 billion in annualized recurring revenue (ARR) from FSD subscriptions alone, assuming steady retention and no major pricing changes.
These figures represent pure, high-margin software revenue. Unlike vehicle sales, which carry production costs, warranty obligations, and supply-chain risks, FSD subscriptions flow largely to the bottom line once the software is developed and deployed over-the-air.
Tesla does not break out exact FSD subscription revenue in its filings (it sits within “Services and Other”), but the category grew 50 percent year-over-year in Q2, with executives highlighting subscriptions as a key driver.
The subscription model offers several structural advantages. It lowers the upfront cost of a new Tesla, potentially broadening the buyer pool and supporting vehicle demand, especially important amid fluctuating EV market conditions. It creates a predictable revenue stream that compounds as the fleet grows and more owners try (and stick with) the software.
Legacy one-time purchasers still exist, but new growth is overwhelmingly subscription-based following the February cutoff.
Early data also suggests improving retention and satisfaction, as well. Tesla has rolled out iterative FSD updates, including v14 features, and expanded availability to additional markets. Recent regulatory approvals in parts of Europe have further boosted interest, with owners in newly enabled countries eager to activate the software they had been waiting for.
FSD is still supervised; regulatory hurdles for true unsupervised autonomy persist in many regions, including the United States, and competition in advanced driver-assistance systems is intensifying. Yet the Q2 numbers validate Tesla’s bet: by removing the large upfront commitment and making FSD accessible via subscription, the company has accelerated adoption faster than many anticipated.
What began as a controversial switch-up has become a clear win. With nearly 1.5 million subscribers, record attach rates, and nearly $1.8 billion in potential annual recurring revenue already in view, Tesla’s FSD business is transitioning from a promised future to a tangible, fast-growing profit engine.
If the momentum continues, and especially if unsupervised capabilities unlock robotaxi opportunities, the subscription flywheel could become one of the most valuable assets in Tesla’s portfolio.
Tesla Robotaxi is among its biggest projects currently, but many have been quick to point out the fact that the company has definitely been slow to expand its fleet.
However, there is definitely a method to that madness. CEO Elon Musk answered several concerns during last night’s quarterly earnings call that some might have about that slow rollout of the Robotaxi suite, maintaining the company’s narrative on prioritizing safety and wanting to avoid injuries to anyone, including animals.
Musk said:
“With Robotaxi, our goals are very ambitious for Robotaxi, but we do need to be cautious about causing any accidents or causing any harm to anyone. Although there are, I think, 30,000 to 40,000 automotive deaths per year in the U.S. alone, most of those do not generate any press or maybe, you never really read about almost any of those. If we injure even one person, it’ll be worldwide headline news, and regulators will immediately clamp down on our activities.
We don’t want to injure anyone. We’re going as fast as humanly possible in scaling Robotaxi, but while trying to ensure that we do not harm anyone at all, and ideally do not even run over a pet. That’s really the constraint is we want to grow as fast as possible with Robotaxi without harm to anyone.”
Tesla has maintained an exemplary safety record with its Robotaxi suite, according to internal data. VP of AI, Ashok Elluswamy, said that the Robotaxi suite has driven more than 380,000 miles unsupervised without any incidents.
0 notable incidents across over 380,000 miles traveled by Robotaxi
— Tesla (@Tesla) July 22, 2026
Analyst Colin Langan of Bank of America also pushed Tesla executives for answers regarding the company’s decision to add cities across several states with dozens of vehicles “as opposed to hundreds.”
Elluswamy said there’s a bigger advantage to do it the way Tesla has been because it ensures that its software stack “is a very general one:”
“The reason we have been expanding across different cities instead of just doubling down on a single city, is that we want to make sure that our stack is a very general one. It is a general one. We just want to both prove to ourselves and to other folks that it is working across a lot of different cities without too much effort per city. That’s what we see internally.”
In the past, we have written about Tesla’s decision to be incredibly conservative with its Robotaxi rollout, especially with the incredibly small fleet size compared to competitors. However, there really is not a price anyone can put on safety for those utilizing the platform or pedestrians, so what Tesla is doing is justified.
A year into the Robotaxi program being active, Tesla has made major strides, but many investors and fans would like to see the fleet expand as quickly as the program has to other cities and states.
Elon Musk
Tesla Semi finally has an FSD timeline and it’s waiting on the Cybercab
Elon Musk told investors Semi self-driving should start working by early 2027, per today’s earnings.
During Wednesday’s’ Tesla Q2 earnings call, an analyst asked Elon Musk when Tesla would look at autonomy for the Semi. His answer set a real timeline for the first time, noting that self-driving on the Tesla Semi is expected to start working “around the end of this year or early next year”.
Musk framed the delay as a matter of priority, not capability. Tesla’s self-driving team is currently focused on Model 3, Model Y, and Cybercab, the vehicles that make up the overwhelming majority of Tesla’s fleet. Since Semi trucks on the road remain a small fraction of that total even after the recent Nevada factory ramp, Musk said it made more sense to keep the software team’s attention on what he called “the march of nines of safety” for the higher volume vehicles first. Autonomous Semi development is “taking a bit of a backseat for the next six months or so,” he said, before adding that it “will definitely be working next year and in time for the scale-up to high production of the Tesla Semi.”
Tesla Semi’s official battery capacity leaked by California regulators
The timeline lines up with what’s already been showing up on public roads. In June, a Tesla Semi was spotted in Sunnyvale wearing a full validation rig, the same rooftop sensor array Tesla mounts on vehicles ahead of an FSD milestone.
A second unit was seen near Fremont days later with a matching camera suite and lens washers. Separately, Tesla analyst Nic Cruz Patane posted video this month of the production Semi’s exterior camera array, ten AI4 based units built directly into the truck rather than added later.
Tesla Semi AI4 cameras. The production version has 10 cameras on its exterior.
These trucks are designed to be autonomous. pic.twitter.com/GH3BamxIBQ
— Nic Cruz Patane (@niccruzpatane) April 14, 2026
Musk also gave the reason autonomy on the Semi matters in the first place, a persistent shortage of qualified truck drivers. “There is a really serious shortage of truckers,” he said on the call, framing a self-driving Semi as important both for addressing that shortage and for improving safety and comfort for the drivers running the truck today.
The timing also tracks with the Semi’s production reality. Tesla’s Q2 shareholder letter, dropped language promising the Semi would reach volume production this year. Musk pointed to 4680 battery cell output as the near-term constraint on Semi and Cybercab production. A software timeline landing in early 2027 gives Tesla’s autonomy team room to work while the hardware ramp catches up behind it.
It’s worth nothing that this isn’t necessarily a promise the Semi ships driverless next year. Musk’s own language, self-driving “working” by early 2027, describes internal validation catching up to hardware already riding on every production truck, not a public unsupervised rollout.
Tesla’s switch-up on selling Full Self-Driving has paid off big time
Tesla Robotaxi’s slow rollout gets explanation from Elon Musk
