TSLA today $411.15 ▲ $2.17 (0.5%) |
| Tesla’s Dojo networking patents move packet recovery into hardware so the training fabric keeps running even when Ethernet drops frames. |
Top 12 News Items
- Goldman Sachs raises Tesla Q2 delivery forecast to 420K: June 16, 2026, TSLAming (X)
Goldman Sachs updated its model after reviewing sales data from the US, Europe, China, South Korea, and Australia. The firm now projects 420,000 Q2 deliveries versus the prior 405,000 estimate and the Street consensus of 400,000. Full-year 2026 deliveries rise slightly to 1.73 million while 2027 and 2028 figures stay at 1.88 million and 1.96 million. The 2026 EPS forecast moves to $1.35 from $1.30 with the price target held at $375 and the rating remaining Neutral. The change reflects stronger-than-expected order trends rather than any shift in long-term assumptions. Source: x.com - Tesla's WO 2024/039800 patent: hardware replay for lossy Dojo network: June 16, 2026, TSLAming (X)
The patent describes a micro-architecture that stores recent packets in on-chip cache or SRAM as small as 256 KB and uses linked lists to track transmission order. When an acknowledgement fails or a timer expires, hardware pointers identify only the missing segment for replay instead of restarting the stream. Four pipeline stages handle stream selection, link identification, replay-or-retire decisions, and pointer updates without software intervention. A hardware link timer working with FIFO memory checks multiple links in round-robin fashion and adapts its tick period as the number of active links changes. This approach keeps Dojo training throughput high even on a controlled but imperfect Ethernet fabric. Source: x.com - Tesla TTPoE: making Ethernet faster by embracing imperfection: June 16, 2026, TSLAming (X)
Tesla’s WO 2024/039794 patent moves transport-layer functions including flow control, acknowledgements, and replay into a TTP MAC block that operates across OSI layers 2–4. Frames carry a specific EtherType such as 0x9AC6 so standard Ethernet switches can forward them while TTP endpoints manage state in hardware. Link state machines transition through closed, open-sent, open-received, open, close-sent, and close-received without TCP-style TIME-WAIT periods. Bounded packet transmission plus local storage and linked-list ordering allow the system to replay only from the first unacknowledged packet. The design accepts occasional loss inside an AI fabric so long as recovery stays fast enough that training jobs never stall. Source: x.com - Tesla Highlights Parallel Parking Feature: June 16, 2026, @Tesla (X)
Tesla posted a reminder that its vehicles can execute parallel parking autonomously. The feature uses ultrasonic sensors and cameras already present on current models to measure the space and steer into it without driver input on the wheel. Owners activate the maneuver through the screen or app once the car has identified a suitable spot. The post underscores that this capability has been available for years yet still surprises some drivers. Source: x.com - Tesla Cybercab with 48 kWh battery and front-mounted motor revealed: June 16, 2026, electrive.com
EPA filings show the Cybercab carries a 48 kWh battery pack and a single front-mounted motor. The documents list estimated range and charging specifications that align with a purpose-built robotaxi rather than a passenger vehicle derivative. The front-drive layout and modest battery size point to lower production cost targets than earlier robotaxi concepts. These details emerged from standard regulatory submissions ahead of any official Tesla release. Source: Google News - Over 100 Tesla Cybercabs Spotted Staging at Giga Texas: June 15, 2026, Not a Tesla App
Drone footage captured more than 100 Cybercab vehicles parked in formation at Gigafactory Texas. The vehicles appear complete and ready for deployment rather than in various build stages. The concentration at one site suggests Tesla is preparing a larger initial fleet than previously visible in public sightings. Production timing and regulatory clearance remain the next milestones to watch. Source: notateslaapp.com - Tesla To Build Pre-Assembled Megachargers for Tesla Semi: June 15, 2026, Not a Tesla App
Tesla is adapting its plug-and-play Supercharger approach to 1.2 MW Megacharger sites for the Semi. Pre-assembled units would reduce on-site construction time and allow faster expansion of high-power depots. The move supports the ongoing Semi rollout by addressing charging infrastructure as a parallel constraint to vehicle production. Details on deployment locations and timelines are still pending. Source: notateslaapp.com - Rivian CEO says company will release tech similar to Tesla FSD later this year: June 15, 2026, Business Insider Africa
Rivian’s CEO stated the company plans to launch a driver-assistance system comparable to Tesla’s Full Self-Driving before the end of 2026. The announcement positions Rivian as a direct software competitor in the same regulatory environment. No technical specifications or naming for the system were provided. The timeline places Rivian roughly alongside Tesla’s push toward unsupervised operation. Source: Google News - SpaceX soars with its first launch as a public company, marking a new era: June 15, 2026, Teslarati
SpaceX completed its first Falcon 9 mission since the IPO on June 15, launching 24 Starlink V2 Mini satellites from Vandenberg. The booster B1093 flew its 14th mission, lifting off at approximately 8:34 a.m. PDT. The flight was described as routine yet symbolically significant for the newly public company. Starlink deployment cadence continues unchanged. Source: teslarati.com - Tesla Robotaxi Fleet Goes 4 Months Without an At-Fault FSD Crash. Humans Keep Hitting It: June 15, 2026, Tesla North
Tesla’s limited robotaxi operations recorded no at-fault collisions over four months of supervised and unsupervised testing. External vehicles struck the test fleet multiple times during the same period. The data covers a small number of vehicles operating in controlled conditions. Regulators will still require larger-scale evidence before approving broader unsupervised deployment. Source: Google News ZuNGxaRTVfeF9pTHREMTdIM3BJZVhCVFVQc1pnZHhhc0I1ZXRabXM0?oc=5 - Tesla (TSLA) Q2 Delivery Projections Exceed Expectations: June 16, 2026, GuruFocus
Multiple analyst notes now project Tesla Q2 deliveries above the original consensus range. The upward revisions follow stronger retail registration data in several markets. The shift narrows the gap between Tesla’s internal expectations and external forecasts. Earnings season will provide the first confirmation of whether the trend holds through June. Source: Google News - Tesla Stock Traded Its Affordable Car Dream for an AI Moonshot. Here Is the Price.: June 16, 2026, Trefis
The analysis examines how Tesla’s capital allocation has shifted from a lower-priced vehicle program toward AI and autonomy investments. It quantifies the implied valuation trade-off between near-term vehicle margins and longer-term robotaxi economics. The piece notes that current share pricing already embeds significant success in the autonomy roadmap. Investors will watch upcoming earnings for any updated capital-expenditure guidance tied to that shift. Source: Google News
Tesla X Takeover: What's Hot Right Now
Tesla X Takeover - What's breaking in the Tesla world today! Here are the most interesting, fresh Tesla developments that have everyone talking.
- EPA and NHTSA terminology needs updating - The post flags that current regulatory language does not yet account for vehicles designed without traditional driver controls.
Tesla’s Cybercab and similar programs will require new definitions around “driver,” “operator,” and responsibility allocation. The observation comes from an account that tracks regulatory filings closely. Updated terminology could affect how future safety standards are written and enforced. Source: x.com - Teslarati plans to buy Cybercab in gold - The account’s editor stated a preference for a gold-finished Cybercab once the vehicle reaches customers.
The comment reflects growing anticipation around the robotaxi’s eventual availability and customization options. It also highlights how the vehicle’s distinctive design is already generating personal ownership interest beyond fleet use. Source: x.com - Tesla’s self-driving safeguards fooled by $30 doll heads - Researchers demonstrated that inexpensive doll heads can trigger incorrect responses in Tesla’s occupant-detection system.
The test raises questions about edge-case robustness in cabin-monitoring cameras and software. Tesla has not yet commented on whether the finding affects current production vehicles or future hardware iterations. Source: Google News - Musk Pegs Subsidies at Under 2% of Tesla, SpaceX Value - Elon Musk stated that government incentives represent less than two percent of the combined enterprise value of Tesla and SpaceX.
The remark came during a discussion of long-term business economics and was framed as evidence that both companies have moved beyond early-stage reliance on policy support. It offers a direct counterpoint to narratives that treat subsidies as a primary valuation driver. Source: Google News - Is Tesla, Inc. (TSLA) A Good Stock To Buy Now? - Insider Monkey aggregates recent institutional filings and sentiment indicators around TSLA.
The report compiles data from multiple large holders without issuing its own price target or rating. It serves as a snapshot of how professional investors have adjusted exposure in the most recent quarter. Source: Google News
Short Spot
Tesla’s self-driving safeguards fooled by $30 doll heads: June 15, 2026, Electrek Researchers showed that low-cost doll heads placed in the cabin can cause Tesla’s occupant-detection system to misclassify the interior. The finding highlights a potential gap in current vision-based monitoring that could affect both safety features and regulatory scrutiny. Tesla has not disclosed any immediate software or hardware response to the test results. Source/Post: Google News
Tesla First Principles
Hardware designers usually treat packet loss as a problem to eliminate through credits, pauses, and lossless fabrics. Tesla’s recent patents instead treat loss as an expected event inside a controlled AI cluster and move recovery into silicon. The approach stores recent packets locally, tracks order with linked lists, and replays only the uncertain segment when a timer or negative acknowledgement arrives. Because the decision logic lives in the chip rather than the kernel, training jobs avoid the latency spikes that software retransmission would introduce.
The economic implication is straightforward: every extra millisecond of stall time across thousands of GPUs compounds into lost training throughput and higher electricity cost per model iteration. By keeping recovery inside the TTP MAC block and adapting the hardware timer to the number of active links, Tesla reduces that overhead without requiring a perfect physical layer. The trade-off is that the system must still guarantee bounded loss rates; an uncontrolled network would overwhelm even fast hardware replay.
This design choice also reveals a broader pattern in Tesla’s AI stack. Rather than buying the most expensive lossless switches, the company is optimizing the entire path from silicon to software for the specific workload of large-scale model training. If the hardware safety net proves reliable at scale, it could lower the capital cost of expanding Dojo relative to peers who rely on traditional data-center fabrics.
Feedback welcome at @teslashortstime. |
