Tether Study Finds Cloud AI Cheaper Than a New Laptop

Tether’s QVAC study estimates that cloud AI can cost less than a new laptop for some workloads, with monthly API fees of $35.75 versus a $50 hardware allocation.

Tether’s QVAC study estimates that cloud AI can cost less than buying a new laptop for some workloads. In its dense-model example, monthly API costs would be about $35.75, compared with a $50 monthly allocation for a $2,399 laptop.

Published October 12, the study spreads the laptop’s cost over four years. It estimates that the dense Qwen3.8-27B model could process about 55 million tokens during a 176-hour working month. Tokens are units of text that an AI model reads and generates.

The example assumes that 94% of the tokens are inputs and 6% are outputs. Alibaba Cloud’s Singapore prices, checked October 22, list the model at $0.50 per million input tokens and $3 per million output tokens. That mix produces an estimated average cost of $0.65 per million tokens.

At 55 million tokens, the monthly cloud bill would be about $35.75. At 22 million tokens, it would be about $14.30. Reaching the $50 hardware allocation would require roughly 77 million tokens per month at those prices.

QVAC measured performance on a 36 GB MacBook Pro with an M5 Max chip, while the $2,399 assumption refers to a separate 32 GB laptop. The comparison assumes that local and cloud models can perform the same work. It excludes electricity, maintenance and any remaining value of the hardware.

The calculation also depends on the model and workload. QVAC estimates that its Qwen3.6 35B mixture-of-experts configuration could process about 200 million tokens during the same 176-hour period. That amount would cover the study’s illustrative 110 million-token coding workload, which is above the dense model’s estimate.

Alibaba Cloud lists different prices for Beijing and charges separately for cached inputs. Lower cache prices or a different mix of input and output tokens would change the calculation. Running a model for more hours could increase the number of processed tokens, provided the work remains within the model’s capabilities.

Tether announced October 20 that it had invested in Tenstorrent, an artificial intelligence computing company. The investment was presented as part of an effort to give developers and businesses more control over computing. Tenstorrent announced October 19 that it had raised $1.5 billion in a Series E funding round at a post-money valuation of $6.2 billion. The size of Tether’s investment was not disclosed.

In an October 21 QVAC account of an October 7 question-and-answer session, Tether Chief Executive Paolo Ardoino said phones do not currently match frontier AI models. He also said that the hardest scientific tasks may remain in large data centers.

Local AI can run a downloaded model without an internet connection and can reduce the amount of data sent to an external provider. QVAC has said that connected tools may still transmit information elsewhere. Neither the Tenstorrent financing nor Tether’s announcement established an integration with QVAC or a reduction in users’ bills.

People who already own a suitable computer do not need to recover the cost of a new device, though electricity and other operating expenses remain. For buyers purchasing hardware mainly for occasional AI use, the device cost continues during periods of low usage while cloud charges vary with consumption.

Articles by this author