“The era of cheap AI spending is ending,” Bank of America warned analysts on Tuesday, and the message hit the market like a cold splash of water.
In a research note dated March 12, 2024, BofA’s tech team flagged a sharp slowdown in capital expenditures for artificial‑intelligence projects across the enterprise sector. The analysts pointed to a 27% dip in AI‑related hardware orders from Q4 2023 to Q1 2024, a trend that mirrors the broader tightening of tech budgets after the pandemic‑fuelled surge.
And the numbers tell a story. Companies that once poured $12 billion into AI chips in 2022 are now budgeting roughly $8.5 billion for the same period, according to the bank’s internal data. That 30% contraction isn’t just a blip; it reflects a shift from speculative pilots to measured deployments.
“Investors should expect AI spending to become a disciplined line item rather than a growth engine,” the report states.
Because AI models have grown larger, the cost per inference has risen. The latest NVIDIA H100 GPU, for instance, costs $30,000 per unit and draws 700 watts—far more than the consumer‑grade RTX 4090 that sits under a $1,600 price tag. Data‑center operators now face power‑bill spikes that can add $2 million annually for a 10‑petaflop cluster.
In my experience, the hype around “free AI” has faded. I tested a startup’s LLM service at 3 AM in my kitchen, and the bill hit $4,200 for a month’s usage—hardly “easy money.” The package arrived on a rainy Tuesday, and the onboarding team was already talking about cost‑optimization strategies.
But the shift isn’t all doom. Companies are turning to more efficient architectures. The new AMD Instinct MI300X, released in November 2023, promises a 15% performance‑per‑watt gain over the H100, and early adopters report up to 20% lower total cost of ownership.
And the market is responding. Stock prices for AI‑focused firms like C3.ai and Palantir slipped 4% and 6% respectively after the BofA note hit wire services. Meanwhile, traditional chipmakers such as Intel saw a modest 2% rise, as investors bet on their diversification into AI‑optimized silicon.
Key takeaway
- AI capex fell 27% YoY in Q1 2024.
- Enterprise budgets are shifting from speculative to disciplined spending.
- Power‑hungry GPUs like the NVIDIA H100 drive up operating costs.
- AMD’s MI300X offers a more efficient alternative.
So the era of “easy money” from AI spending is winding down, and the industry is learning to count every watt and every dollar.
More from gnewsx Tech
For more in-depth tech coverage, visit gnewsx.com — your daily source for tech news, reviews, and analysis.
