“It is important to note that Tesla is by far the best in the world at real-world AI,” Elon Musk told analysts in the past year. Nevertheless, in view of the optimistic forecast, he has already changed his estimate for creating artificial general intelligence, this time pushing the date for 2026, having previously projected it for 2025. This indicates how volatile estimates in creating artificial general intelligence can be, where definitions are highly disputable and challenges are so enormous.
AGI, as described by IBM, would require AI to equal or surpass human intelligence in all tasks, whereas today’s the best and largest language models are still based on statistical pattern recognition and haven’t attained true reasoning capabilities, as AGI implies. A similar vision of scaled infrastructure is pursued by Musk’s xAI, which is now planning what it calls “Colossus,” starting from just 200,000 graphics processing units (GPUs) and planning to rise to one million. Such an upgrade is actually the heart of this short-time-line hypothesis because, over the years, increased computation has been associated with breakthroughs in AI, and predictions indicate that 10,000× GPT‑4’s training compute power could be within reach by 2030.
In his company-wide meeting recently, Musk targeted his 2026 AGI deadline to when Grok 5 is being launched by xAI. The capitalization pipeline in his company at $20 billion to 30 billion annually was unprecedented in his biggest rivals in developing AGI, which are OpenAI and Google DeepMind. His move to incorporate Grok in Teslas earlier this year is part of his strategy to combine his companies in his portfolio strategy to quickly roll out AGI technologies. This company event also showcased Grok Voice’s improvements in listening and prediction capabilities in optimizing AGI.
Whether the above scaling alone will enable the achievement of AGI remains a topic of concern for technical experts. The proponents of the short timeline discuss the fast saturation of research benchmarks in the realm of MMLU and GPQA, where the current LPFs approach or may surpass the results of human experts. They notice that the AI systems are performing longer and longer tasks, and some tests even indicate the doubling of the ability to perform tasks for longer periods every few months, so the autonomous projects lasting for months may become possible within the decade.
Skeptics have argued that the benchmarks merely measure a limited set of skills, while more realistic scenarios would necessarily be prone to fuzzy objectives, ambiguous dependency relations, and sensor-motor interface complexities that have proved daunting even for humans. The irony pointed out by Par Moravec’s Paradox is that abstract reasoning may be more easily captured by machine logic than the dexterous activities involved in routine physical labor. This is especially applicable with Musk’s secondary project that is, the volume production of Tesla’s Optimus, a biped humanoid. While it is intended for use in both factory work and household chores, including taking care of children, the currently available models are necessarily controlled remotely for their more intricate activities. While Boston Dynamics’ and Agility Robotics’ rivals’ product lines do feature interesting advances in their biped designs, further innovations would be required for both mechanical sophistication and low-power operation if mass production is envisioned.
Musk’s AGI story also relates to another area of research engineering speculation, such as extraterrestrial data centers. Musk has mentioned the possibility of space-based Optimus data centers, which would not only need research and development work regarding self-replication and maintenance by autonomous machines, as well as radiation hardening and space logistics, but Musk’s willingness to connect such space data centers to the timeline of AGI development shows how research engineering ambition can entangle with marketing.
A lack of resolution on what defines AGI portends to making progress very difficult to track. Some scientists have emphasized economic utility, along the lines of doing tasks more efficiently than humans do in most economically significant tasks. Others focus more upon self-awareness and consciousness. Without any clear consensus, crossing such milestones might be potentially called upon prematurely or overlooked completely, depending on criteria. It would fuel the debate and skepticism, and Musk’s ever-shifting timeline has been one way to see just how much the public debate about AGI has come to muddy the waters between the lines of what is technically possible and simply market positioning among competitors.
As to investment and market analysis, the ramifications are clear: if xAI’s approach to scaling pays off, it has the potential to be incredibly significant in every industry. But if not, another adjustment to expectations in that sector might be what’s forthcoming, and it would continue the very long cycle of predictions about AGI delays just around the corner.
