The software industry is standing before a major financial turning point as the operational costs of artificial intelligence (AI) systems tend to rise faster than personnel costs. According to the latest analysis by venture capitalist Tom Tunguz shared on Hacker News, the hardware and computing costs to maintain large-scale AI models could surpass the average salary of a software engineer by 2029. This trend forces technology businesses to reassess their operational cost structures and long-term business models.
Background & Causes
For decades, the largest cost for most software companies has always been the payroll for engineers and developers. However, the boom of large language models (LLMs) and generative AI systems has completely changed this financial equation. Training and maintaining large-scale AI models require massive computing resources, accompanied by skyrocketing investments in specialized hardware like GPUs and electricity bills continuously rising over the years.
Technical Analysis & Technology
To better understand this forecast, we need to look at the growth rate of parameters in AI models compared to Moore's Law. While semiconductor performance improves in a linear cycle, the size of AI models is growing exponentially. Although the inference cost per query for frontier models has decreased due to algorithmic optimization, the total volume of queries from millions of concurrent active users has multiplied, placing immense financial pressure on cloud infrastructure.
Expert Opinions & Remarks
According to Tom Tunguz, the breakeven point where the AI operational cost per engineer equals their actual salary will be a milestone that reshapes the tech industry. Many experts on Hacker News also expressed agreement and noted that startups failing to optimize model performance will soon face cash flow depletion. Conversely, some more pragmatic views suggest that the development of cheap custom processors and model distillation techniques could help delay this milestone.
Impact & Future
This shift in cost structure will strongly impact how tech companies allocate budgets in the future. Instead of solely focusing on hiring high-quality personnel, companies will have to invest more in optimizing infrastructure and choosing small-scale but highly efficient models (SLMs). For the tech community in Vietnam, this is an opportunity for engineers to dive deep into performance optimization and cloud cost management (FinOps), skills projected to be extremely valuable in the next AI era.