The Alpamayo 2 Super Mirage: NVIDIA, Robotaxi Hype, and the Machine That Prices Nothing

The headline crossed my terminal at 03:47 Tokyo time. NVIDIA had released something called Alpamayo 2 Super. An open AI model for commercial Robotaxi development. By 04:10, AI tokens were up. NVDA futures ticked higher. By 06:00, I had checked the official NVIDIA page. Nothing. No blog post. No model card. No GTC slide. Just a story on a crypto media outlet. The market doesn't wait for verification. It waits for liquidity. This is the trade that made my morning. I have spent years watching this dance. A headline appears. Retail chases. Smart money uses the spike to exit or accumulate. The product might be real. It might be vapor. The truth is secondary to the timing. Alpamayo 2 Super is the perfect stress test for anyone who claims to trade information. Here is what we actually know. Almost nothing. Two data points, both sourced from a single article. The model supposedly targets development, not production. It is supposed to support inference, planning, and training. That is it. No architecture. No parameter count. No license terms. No official NVIDIA confirmation. The analysis below is built on inference, not evidence. That is exactly how this market works. The first thing serious traders do when a phantom appears is check the trail. The trail starts with context. NVIDIA already told us this was coming. At CES 2025, the company announced NVIDIA DRIVE AI with a model called Alpamayo. The name comes from a mountain in Peru, part of the Cordillera Blanca range. In NVIDIA's vocabulary, that mountain became a foundation model for autonomous vehicle development. DRIVE AI is the orchestration layer. Alpamayo is the brain. So when a report says Alpamayo 2 Super, the name is not absurd. It fits a product sequence NVIDIA established months ago. The second generation, with a Super suffix, would be a performance bump. That is consistent with how NVIDIA labels hardware. It has used Super as a marker of enhanced capability for years. Plausible, yes. Confirmed, no. The publishing source matters. Crypto Briefing is a Web3 media outlet, not an automotive trade publication. That does not make a story false. It does make a story thin. The original report contains no interview, no technical white paper, no NVIDIA quote, and no official link. It appears to be a fast summary of an event that might have been a leak, a press release that was later pulled, or a misunderstanding of an existing roadmap. In a market where information is the product, thin sourcing is a liability. I treat it as untradeable until confirmed. But I still analyze it, because the structure is informative even if the specific news is not. The structure is where the real work lives. So let me rebuild the entire picture from the ground up. Seven layers. Seven ways to understand why a phantom product can still move money. And why the smart play is often to do nothing while the crowd does everything. Layer one is technical reality. Alpamayo 2 Super, if it exists, is not a full autonomous driving system. It is a model layer. The phrase open model, applied to a driving AI, means one of a few things. It could be a vision-language-action model that takes camera and sensor data, understands the scene, and outputs control decisions. It could be a world model that predicts future states of the road environment for simulation and planning. It could be a simulator-focused model inside NVIDIA's Omniverse stack. The original report says the model supports inference, planning, and training. That combination points to a foundation model rather than a production-grade L4 stack. NVIDIA is not trying to replace Waymo's full driver. It is trying to give every OEM and startup a better starting line. What would the architecture look like? Almost certainly a transformer-based backbone. A vision encoder for cameras. A text encoder for map and instruction data. Some form of cross-attention to fuse the sensor streams. And a head that either outputs a trajectory or a set of semantic predictions. The parameter count is unknown, but a serious driving model in the current era sits in the tens of billions of parameters. Not a hundred billion. Not a trillion. Enough to be useful, small enough to be compressible. That scale is not a guess. It is the economic logic of the application. You need enough capacity to handle edge cases, but you also need a path to a vehicle-grade chip that consumes less than a few hundred watts. The Super suffix implies an iteration. First generation Alpamayo was announced as part of NVIDIA DRIVE AI. A second generation would improve inference speed, planning accuracy, training efficiency, or all three. Whether that improvement is real or marketing is unknown. NVIDIA has a history of using Super as a real differentiator in performance. In the chip world, a Super version typically means more memory bandwidth, higher clocks, or better power efficiency. In a neural network, Super could mean a distilled model that runs faster on DRIVE Thor. It could also mean a larger model that needs a data center to breathe. The name alone cannot tell us. I don't trade names; I trade structure. There is another classification problem hiding in the announcement. The report says the model supports training. That is unusual for a vehicle-level model. Autonomous driving systems do not typically train themselves inside the car. A model card that supports training is a data center product. It is something a research team runs on DGX hardware, not something a Robotaxi runs on the road. So the so-called open model probably has two faces. One face is a pre-trained foundation model that developers use for research. The other face is a distilled deployment model that customers fine-tune and push to the vehicle. The same brand name covers both. This is not deception. It is how AI products are actually built in 2026. But it creates confusion for anyone who reads the headline as a single artifact. That confusion is the entry point for bad trading. A trader needs to know what is being announced. A developer needs to know how to run it. A regulator needs to know what it is allowed to do. The original report answers none of those questions. It is a marketing echo, not a technical document. When I audited token sale contracts in 2017, I read every line of the smart contract before I signed anything. I found three reentrancy vulnerabilities that could have drained four million dollars. The client was upset. The contract was delayed. But the funds were protected. That lesson never left me. Read the actual artifact. If there is no artifact, your analysis is a hypothesis. And a hypothesis is not a position. The technical certainty level here is low to medium. The direction of NVIDIA's platform strategy is well documented. The specific claims about Alpamayo 2 Super are not. That distinction is important because it changes how you size any trade. A confirmed product with a model card is a different risk profile than an unconfirmed rumor. In the absence of confirmation, I keep the analysis, I keep the thesis, and I keep my war chest untouched. Layer two is commercial logic. Alpamayo 2 Super is not a product. It is a funnel. NVIDIA makes money when the model pushes developers deeper into its ecosystem. The model may be open, but the ecosystem is not. To train a custom version, you need DGX compute or DGX Cloud. To validate it, you need Omniverse simulation. To run it in the vehicle, you need DRIVE Thor or DRIVE Orin. To do all of this efficiently, you need CUDA. That is the business model. The model is the hook. The hardware and platform are the margin. The market doesn't pay for models. It pays for lock-in. Let me make the commercial chain explicit. A mobility company reads the headline. They request access. NVIDIA gives them a model card and a sample notebook. The notebook runs on a DGX reference architecture. The company starts fine-tuning on NVIDIA Cloud. Every iteration is logged in an NVIDIA environment. The validation suite uses Isaac Sim scenarios. The final vehicle integration uses DRIVE Thor. The company thinks it is using an open model. In reality, it has entered a commercial orbit. The exit cost becomes prohibitive. That is not an accident. That is design. Open has a wide semantic range. An open model can mean public weights, a permissive license, or a developer program behind a wall. The word open does not mean free. It does not mean commercially unrestricted. It does not mean the training data is public. Every serious trader knows this trick. The phrase open model is a marketing asset, not a technical guarantee. NVIDIA has every incentive to make the model easy to test and hard to leave. Weights might be downloadable. Fine-tuning might require an approved NVIDIA cloud account. Deployment might require DRIVE hardware. The lock is not in the file. The lock is in the toolchain. This is consistent with NVIDIA's broader strategy. They are no longer just the chip company. They sell AI foundries, AI factories, and now AI models. The automotive vertical combines DRIVE Thor, DRIVE OS, Isaac Sim, Cosmos world models, and DGX infrastructure. Alpamayo 2 Super would be the model that binds these pieces together. Its commercial value is indirect, but massive. Every developer who adopts it is more likely to buy the full stack. Every OEM that uses it is more likely to design around a DRIVE Thor board. The model is a loss leader with a razor inside. There is a hidden boundary the report ignores. The model may be open to approved partners only. NVIDIA has reason to restrict access. Regulatory pressure, export controls, safety liability, and commercial exclusivity all point to a guarded release. A truly public weight dump would create compliance headaches. NVIDIA knows this. That is why the phrase commercial Robotaxi development is in the announcement. Development access can be granted to companies that sign agreements. Production access can be controlled by hardware certification. The word open becomes a permission structure, not a public release. Investors should not make the mistake of valuing Alpamayo 2 Super as a standalone revenue stream. It is not. It is a customer acquisition mechanism. The revenue lands in the data center segment, the automotive segment, and the cloud segment. NVIDIA does not need to sell the model. It needs the model to sell everything else. That subtlety is lost on most media coverage. And that lost subtlety is an edge. Layer three is industry impact. A credible open model would reshape the Robotaxi race. Consider the barrier structure right now. A serious autonomous vehicle program needs data collection fleets, expensive compute, simulation pipelines, and teams of researchers. The base model is the hardest part to build. An open model supplied by NVIDIA removes that barrier. A mid-tier mobility company can start from Alpamayo 2 Super instead of hiring two hundred researchers and collecting a million miles of edge cases. That accelerates the timeline by years, not months. This is the democratization narrative. But it is half true. Democratization cuts two ways. The same model that lowers the floor also raises the ceiling. NVIDIA controls the model, the compute, the simulation environment, and the hardware certification path. Every customer is a tenant, not an owner. The open model makes NVIDIA the standard setter. It strengthens NVIDIA's bargaining power against chip competitors like Qualcomm and Mobileye, because those companies can no longer match the full stack. It also pressures Chinese alternatives. If Alpamayo 2 Super is restricted by US export rules, the Chinese market becomes a parallel universe where domestic chips and domestic models must fill the gap. That gap is a geopolitical trade, not a technical one. And trade, in this environment, is an opportunity. The open-source precedent is real. Open language models changed the application layer of AI. Instead of every startup training a model from scratch, they fine-tuned Llama or Mistral. The same dynamic could hit autonomous driving. A company like a ride-hailing operator in Southeast Asia does not want to build a foundation model. It wants to run a service. An open driving model gives it a starting point. Fine-tuning with local road data becomes feasible. The long-tail scenarios, the unusual traffic rules, the local signage, all of that becomes a fine-tuning problem instead of a fundamental research problem. This is genuinely transformative. But there is a structural trap. The fine-tuning data is the asset. The local operator generates valuable data during adaptation. That data flows through NVIDIA's toolchain. NVIDIA can see what its customers are struggling with. It can update the base model to fix the general cases. Over time, the base model grows stronger because every customer is unknowingly contributing to the ecosystem. The customer receives a better starting model. NVIDIA receives the most valuable intelligence in the world: a map of every failure mode across every market. That intelligence is worth more than any chip sale. And it never appears on a balance sheet. Who benefits from this change? Data annotation companies benefit. Simulation testers benefit. Robotaxi operators benefit. They all get cheaper access to a better starting line. Who loses? Full-stack startups that built their own foundation model at great cost. They now have an alternative that is good enough and infinitely cheaper. Their research advantage becomes a cost disadvantage. The market will reprice them accordingly. There is nothing moral about this. It is simply the economics of a model layer commoditizing the lowest level of the stack. What about the established leaders? Waymo has its own data engine. Tesla has a fleet that collects millions of miles every day. They do not need Alpamayo 2 Super. But they will watch it carefully. If NVIDIA's model makes every mid-tier competitor competent, the competitive floor rises. That is good for nobody who wants to retain a proprietary edge. It is, however, good for NVIDIA. The company becomes the referee in a game it does not play. Layer four is competition. Let me lay out the players as they deserve. Waymo is vertically integrated. It builds its own hardware, writes its own models, and operates its own service. An open NVIDIA model is a compliment to that strategy, not a threat. Waymo does not need to buy Alpamayo 2 Super, because its own model is more specialized. Tesla is similar. Tesla has fleet data that no one else has, and it deploys on its own Full Self-Driving computer. NVIDIA can sell Tesla training GPUs, and it does. But the open model will not appear in a Tesla dashboard. The real fight is for everyone else. Mobileye and Qualcomm sell chips to the same OEMs that NVIDIA is courting. If Alpamayo 2 Super makes NVIDIA's silicon the easiest path to an L4 prototype, those chip rivals lose the top of the market. They can keep the L2 business. They can fight for the middle. But the high-end Robotaxi story will default to whoever owns the model. That is the quiet war. It is not about TOPS or watts per inference. It is about which architecture has the best model available on day one. NVIDIA is not just selling a chip. It is selling the brain that makes the chip useful. NVIDIA already dominates the L3 and above development market. The estimate in the source material sits above eighty percent. That number is hard to verify, but the signal is clear. NVIDIA's CUDA ecosystem, its simulation suite, and its data center dominance make it the default choice for serious autonomous driving research. Alpamayo 2 Super, if real, extends that dominance into the model layer. It makes NVIDIA the only supplier that can claim a full stack from training to edge inference. That is a powerful message to OEMs who want one throat to choke. The Chinese picture is more complicated. Horizon Robotics and Huawei have domestic alternatives. If NVIDIA's model cannot legally cross the border, those domestic alternatives gain a protective moat. Chinese Robotaxi leaders like Baidu Apollo, Pony.ai, and WeRide will not wait for NVIDIA's export license. They will build their own foundation models. That is why the most important competitive question is not whether Alpamayo 2 Super is good. The question is whether it can be deployed in China. If no, then NVIDIA has accidentally subsidized the Chinese competition by making every Chinese executive realize they need a domestic model stack. The market will price that accordingly. US export controls are the wildcard. AI model weights are increasingly treated as dual-use technology. A foundation model trained for autonomous driving has military applications. Drones, unmanned ground vehicles, logistics, all of it. The US government has debated restricting AI model weights for years. If Alpamayo 2 Super contains state-of-the-art perception and planning capability, it could fall under export restrictions. That would limit its international adoption. It would also create a black market for the weights if they leak. The export control question is not a footnote. It is a structural determinant of who wins the global Robotaxi race. There is also a brand anchoring effect. The name Alpamayo 2 Super sounds like a leap forward. It is probably an incremental update. Media outlets love the Super suffix because it generates clicks. NVIDIA understands this. It has used naming as a marketing weapon for decades. A trader who sees the name and assumes a revolution is making a category error. The real strategic shift happened years ago, when NVIDIA decided to become an AI platform company. Alpamayo 2 Super is just another step on that road. The market doesn't sell you the truth. It sells you a position. Layer five is ethics and safety. I will be blunt. Open sourcing a driving model is not like open sourcing a language model. A hallucinating chatbot wastes your time. A hallucinating planning model kills a pedestrian. The automotive world has strict standards: ISO 26262 for functional safety, ISO 21448 for safety of the intended functionality. An open model from NVIDIA will not come with a safety case. It will come with a disclaimer. The customer assumes the risk. This is not a criticism of NVIDIA. It is the structure of the industry. But it matters when media reports call the model a turning point. The development use case is safe enough. The production use case is a legal and ethical minefield. A company that deploys a model fine-tuned on Alpamayo 2 Super into a real vehicle takes on the liability. If the vehicle crashes, the OEM is responsible. NVIDIA has no skin in that game. Its contracts will be written to leave functional safety with the buyer. That is standard in the industry. But it is worth saying out loud because the phrase open model sounds dangerously casual. There is no such thing as a casual deployment of a four-thousand-pound machine. The original report does not address this. No mention of safety validation, no mention of responsibility, no mention of data privacy. That absence is itself a signal. NVIDIA is positioning this as a developer tool, not an approved component. In my audit work back in 2017, I learned to read what a contract does not say. The missing warranty is the most important clause. The same applies to Alpamayo 2 Super. If the model supports commercial Robotaxi development, the final safety burden is on the OEM. I value that clarity. It means the open model is not a product guarantee. It is educational material for a dangerous machine. Data privacy is another layer. Robotaxi models ingest thousands of hours of camera footage. That footage contains faces, license plates, and movement patterns. Training data must respect GDPR in Europe, China's Data Security Law, and a dozen other regulations. An open model with released weights could leak private information. The original report gives no path for compliance. That is a red flag for a serious buyer. It is not necessarily a dealbreaker. It is a due diligence item that must be prioritized before deployment. I don't rely on press releases for compliance. I rely on lawyers. There is also the dual-use problem. An open driving model can be adapted for military unmanned vehicles. The same perception stack that avoids a pedestrian can avoid a patrol. The same planning network that navigates a city can navigate a supply route. NVIDIA is well aware of this. The company has to balance its commercial ambitions against export rules and national security restrictions. That is why the announcement is so carefully worded. Commercial Robotaxi development is a clean, civilian framing. It does not mention drones or defense. But the technology does not care about the framing. The market should not ignore the downstream risk. What is the confidence level for the safety analysis? Low to middle. It is based on general patterns in the autonomous vehicle industry, not on specific details of Alpamayo 2 Super. There is no official safety document. There is no functional safety certification. There is no test report. The absence of those artifacts is the only data point. I treat the responsibility gap as the default assumption until proven otherwise. That assumption has saved me money many times. It will save you money too. Layer six is investment. This is where the crypto market and the stock market collide. Alpamayo 2 Super, real or not, is a narrative object. It sells a vision of NVIDIA as a full-stack AI infrastructure provider. That vision supports the long-term valuation story around AI factories, robotics, and autonomous fleets. It does not add a new revenue line by itself. NVIDIA's revenue is dominated by data center GPUs. Autonomous driving is a small percentage of the total. A single model announcement, even a real one, will not move the quarterly earnings line. But it will move sentiment. And sentiment, in this market, moves faster than fundamentals. The typical mistake is to read this as NVIDIA entering the Robotaxi operating business. That is wrong. NVIDIA is not going to build a ride-hailing fleet. NVIDIA is going to sell the tools that make ride-hailing fleets possible. The difference is massive. A journalist writing about an open model creates a mental image of an autonomous car. A trader sees a data center order. The market narrative oversimplifies the economics. Smart money sells the simplification. That is where the edge lives. There is also a second-order effect on crypto. AI-related tokens cluster around NVIDIA announcements. A headline about Alpamayo 2 Super can trigger a short-term pump in coins with names tied to autonomous driving, robotics, or AI agents. Usually those pumps fade. If the underlying product is unconfirmed, the fade is fast and brutal. I have seen this pattern enough times to treat it as a calendar event, not an investment thesis. Buy the rumor, sell the news? No. In this case, the rumor was a ghost. There is no reliable news to sell. So the proper trade is to wait. Wait for the official announcement. Wait for the model card. Wait for the latency and parameter counts. Then decide. The valuation impact on NVIDIA itself is indirect. A strong automotive AI story supports the long-term growth narrative. It gives sell-side analysts another reason to extend their target dates. It creates positive sentiment carry into the next earnings call. But it does not change free cash flow today. It does not change the data center sales cycle. It does not change the gross margin profile. Any investor who marks NVIDIA up by ten billion dollars because of an unconfirmed model announcement is projecting, not calculating. And projections have a way of reverting to reality. The market has a weak memory for phantom products. I remember the DeFi summer of 2020. I deployed capital into yield farming with real positions, not paper models. I learned what an oracle manipulation feels like when a $12,000 liquidation hit my account. That pain taught me that on-chain mechanics behave differently under stress. The same is true for corporate announcements. A headline behaves differently under stress. A phantom product can disappear. When the story dies, the price reverts. If you bought the spike, you become the exit liquidity. The market doesn't reward the fastest fingers. It rewards the cleanest hands. Layer seven is infrastructure. If Alpamayo 2 Super exists, it needs a home. The model is probably far too large for a single vehicle-grade chip. Training a foundation model for autonomous driving requires thousands of GPUs. The original report mentions training, inference, and planning as supported tasks. Training happens in a data center. Inference and planning happen in the vehicle. That creates a two-sided demand story. On one side, NVIDIA sells DGX SuperPODs and H200 or GB200 clusters for the training phase. On the other side, NVIDIA sells DRIVE Thor for the edge deployment. The model is the bridge between those two revenue streams. That is a much more powerful business than selling a model alone. Infrastructure has a hard constraint. A large driving model is expensive to run in a car. Every watt of compute shortens the electric vehicle range. Every millisecond of latency reduces the safety margin. Production deployments will need distillation and quantization. The model will be compressed to fit the chip. NVIDIA knows this. The Super version might already be the distilled one, or it might be the research model that customers distill. Either way, the winner is NVIDIA. Because the physical boundary of the vehicle forces the customer to stay inside NVIDIA's hardware ecosystem. The model requires the chip, and the chip requires the model. That is the deepest lock-in the company owns. Let me be specific about the compute story. A serious driving foundation model trains on clusters of thousands of GPUs. The training run can consume months of time, even with a well-optimized cluster. That is a DGX SuperPOD sale. It is also a software contract. It is also a power and cooling contract. The model is a load generator for the entire data center sales pipeline. Every AI factory that NVIDIA has built with partners becomes a potential training site. The company does not need to release the model globally. It needs to release it where it can capture the compute spend. That is the strategy behind the so-called AI factories with partners like Alibaba Cloud and Aston Martin. Edge deployment is the second half. DRIVE Thor is the current flagship vehicle computer. It is designed to run large neural networks at automotive power levels. But a foundation model with tens of billions of parameters cannot run as-is. It needs pruning, distillation, quantization, and operator fusion. That work is expensive and difficult. NVIDIA will offer the tooling to make it easier. It will also sell the hardware that makes it possible. The result is a beautiful economic machine. Each graduate from the research stack becomes a hardware customer for the deployment stack. The open model could even be a treadmill. If the public version is the research-grade model, customers spend money to make it deployable. Then NVIDIA releases an improved version. The cycle repeats. This is not a conspiracy. It is a product strategy. Every technology platform that survives does this. The key insight is that the value is not in the weights. It is in the ability to move through the training-to-deployment cycle faster than anyone else. NVIDIA has made that cycle its own. Now the contrarian angle. The crowd sees an open model as a gift. I see a long-term consolidation of power. But there is an even sharper blind spot. The open model could be too open. If the weights are truly public and the license is permissive, a determined competitor can run the model on a different GPU architecture. CUDA is a moat, but not an absolute moat. Open standards and translation layers have eroded such moats before. If Alpamayo 2 Super runs on hardware that is not NVIDIA's, the model becomes a commodity. NVIDIA loses the exclusivity that makes the entire strategy work. Why would NVIDIA risk that? It probably will not. The open label will be carefully managed. The company has every incentive to keep the full toolchain proprietary. That is why I believe NVIDIA will never make the model truly portable. It will open the weights. It will not open the toolchain. And if someone finds a way to run it elsewhere, the company will update the terms or the model. The market doesn't sell you the truth. It sells you a position. There is a second blind spot. The original source might be a misunderstanding. I have seen more than one phantom product in this industry. A careless media piece takes a slide from an old presentation, adds a Super suffix, and invents a headline. The market reacts. The official channels say nothing. The story slowly dies. The most dangerous position is the one you take on a headline you cannot verify. My rule, hard learned after the 2020 DeFi oracle attack, is to never enter a position where the liquidation threshold is in someone else's hands. A phantom product is exactly that. The price action is real, but the cause can vanish. The Terra collapse in 2022 reinforced this. I held no large stablecoin exposure because I refused to rely on a single protocol. Everyone thought I was being paranoid. Then the protocol collapsed, and my preservation happened to look like genius. It was not genius. It was a rule. The rule was simple: if I cannot verify the foundation, I do not build on it. Alpamayo 2 Super is an unverified foundation right now. The strategic analysis is valid because the pattern is real. The specific trade is not. There is a third contrarian layer. If the model is open and the weights leak, Chinese teams could adapt it quickly for domestic chips. That would give NVIDIA zero hardware revenue on those deployments. It would also accelerate the Chinese autonomous driving industry. In that scenario, NVIDIA creates its own competitor. The company will try to prevent this with export controls and license agreements. But software leaks. Weights leak. The history of open source is full of runaway copies. The market does not price this risk well because the market is staring at the upside of the announcement. The contrarian sees the downside of the leak. What about the possibility that the announcement is a deliberate leak? NVIDIA may be testing market response. Companies do this all the time. A carefully planted story generates a market signal without the legal commitment of a press release. The executive team watches the ticker, watches the commentary, and decides whether to move forward. If the reception is positive, the official announcement comes next week. If the reception is confused, the story is quietly buried. This is the shadow launch. It costs nothing and provides instant feedback. I have participated in these dynamics on the analytics side. The pattern is always the same: first a whisper, then a wave, then a formal document. If you learn to read whisper to wave, you can position before the formal document. But you must be prepared for the wave to die. Let me give you the signals to track. In the short term, watch for an official NVIDIA blog post or developer forum creation. If the model exists, there will be a model card on a public registry. If there is no official announcement within one or two weeks, treat the original story as a false positive. In the medium term, watch GTC and CES. NVIDIA uses those stages for real launches. In the long term, watch the quarterly earnings for DRIVE-related revenue. That number will tell you more than a press release ever will. Here is the actionable part. Do not buy the first spike. If Alpamayo 2 Super is real, NVIDIA will repeat the story with a full technical release. At that point, you can evaluate the parameter count, the license, and the hardware requirements. If it is vapor, you avoided a corpse. The tradeable opportunity is not the headline. It is the confirmation. That is true in equities, and it is true in crypto. A confirmed event changes the playbook. If NVIDIA officially releases the model, I would look at three things. The first is the license. Permissive weights with a restrictive toolchain is a bull signal for NVIDIA. Restrictive weights with an open toolchain is a bear signal for the moat. The second is the hardware requirement. If the model only runs on DRIVE Thor, then every Robotaxi developer becomes a hardware customer. The third is the pricing. If NVIDIA charges for inference credits or cloud fine-tuning, the model is a direct revenue stream. I don't need to know the price of every stock. I need to know the price of every dependency. What is my actual prediction? I expect that NVIDIA will release some form of Alpamayo iteration in the next product cycle. The autonomous vehicle strategy demands it. The model will be called something close to this, and it will be integrated with DRIVE Thor and Cosmos. But I also expect the open label to be a carefully managed enclosure, not a true open source release. The lock-in is the product. I don't expect NVIDIA to compete with Waymo on the road. I expect NVIDIA to make everyone else less dependent on Waymo by offering a standard model, and then charge for the privilege of actually using it. The deepest trade here is not the model. It is the realization that autonomous driving is becoming an AI chip business first and a vehicle business second. Every vehicle maker that uses an NVIDIA model is capping its own intellectual property. The value migrates upward to the platform provider. In the same way that mobile phone profits migrated to the chip and OS layer, Robotaxi profits will migrate to the model and compute layer. That is the long-term thesis. It survives even if Alpamayo 2 Super is a fiction. The story will be told with a different name. I have been doing this since the 2017 ICO audits, when I refused to sign off on a contract that was vulnerable to reentrancy. The client was unhappy because the fix meant a delay. But the fix also meant the funds would not be drained. I did not care about the client's timeline. I cared about the code. That same refusal to bend to social pressure is why I survived 2020's oracle manipulation with a loss I could afford, why I held no Terra stablecoins in 2022, and why I am writing this now without pretending that the facts are solid. I don't write certainty where only inference lives. So what do you do with Alpamayo 2 Super? Treat it as a map, not a destination. The map shows you where NVIDIA is going. It shows you the shape of the competitive war. It shows you which losers in the Robotaxi race will be rescued by a rented brain. It also shows you the exact point where media fantasy meets market liquidity. The only thing it does not show you is a valid entry price. The price will come later. Here is my final rule. The market doesn't care about your opinion. It cares about your position, your size, and your exit. If you cannot name your exit before you enter, you are not trading. You are gambling. And gambling on a headline from a crypto outlet about a model NVIDIA has not announced is a special kind of self-harm. I am not asking you to ignore the news. I am asking you to time the confirmation. There is a difference between knowing a waterfall exists and standing under it. Wait until the structure is clear, then position. The Alpamayo 2 Super story, whether real or rumor, will teach you that lesson if you let it. And if the official announcement never comes, you will have learned it for free. The market doesn't give away many gifts. This one is a free education. Do you want to trade a ghost, or do you want to trade the confirmation that actually lands? The distinction is the whole game. I don't need to be first. I need to be right. And right means waiting for the structure to prove itself before I lock in a risk that someone else invented.