Businesses Seek Cheaper AI Models Amid Soaring Costs

tech360.tv
1 day ago
3 min read

Businesses are now considering more affordable artificial intelligence models. This shift follows rising expenditure on powerful AI tools, prompting a re-evaluation among tech executives. High costs, driven by unpredictable usage based pricing, have led some organisations to cap internal AI use.

Leading executives, including Microsoft's Satya Nadella, Palo Alto Networks' Nikesh Arora, and Coinbase Global's Brian Armstrong, suggest smaller, cheaper models can fulfil significant corporate needs, according to Reuters. This perspective emerged from companies reassessing AI tool consumption, a practice termed "tokenmaxxing".

Prices for AI tokens, the units measuring usage, are decreasing. But the total cost of completing a task often rises. AI firms are shifting from flat subscriptions to usage based pricing, leading to unpredictable, higher bills. Estimating usage per task has become more challenging.

Uber, for instance, exhausted its entire 2026 AI budget in four months after employees rapidly adopted AI coding tools, forcing management to impose usage limits. Harold Byun, CEO of BlueRock, reported customer spikes of 20 per cent to 30 per cent in over budgeting due to licence model changes.

As companies increase AI reliance, costs frequently exceed initial estimates, with tasks demanding more steps, data, and longer inputs. Gartner predicts AI coding expenses will surpass the average developer's salary by 2028. A survey found three quarters of executives foresee rising tech budgets, nearly half projecting double digit percentage increases.

Businesses now employ routing tools such as OpenRouter, an AI marketplace. This assigns tasks to the most cost effective system, reserving premium models for complex work. And open source tokens processed on OpenRouter increased to 65 per cent recently, from 34 per cent earlier, a Citi note indicated.

This trend benefits open source model makers like China's DeepSeek, which gained wide adoption among startups but struggled to enter larger businesses due to security concerns. DeepSeek holds the top position among the four most popular Chinese models on OpenRouter.

Palo Alto Network's Arora recently urged AI labs to price tokens at anticipated lower future rates to attract enterprise clients. OpenAI, ChatGPT's creator, is reportedly considering substantial price cuts, including for token usage, ahead of similar expected actions from rival Anthropic.

However, cheaper models could negatively affect revenue growth, especially as these companies prepare for potential IPOs. But a "price war dynamic" is anticipated between OpenAI and Anthropic as they compete for initial public market dates, according to Christopher Brown, a financial adviser at Synovus Securities.

Tech stocks experienced a sell off recently. Investors reassessed AI valuations following doubts about returns on extensive spending. Weak post IPO performance from SpaceX and reports of a potential delay in OpenAI's listing compounded these concerns among Big Tech investors.

Chinese models are closing the capability gap with leading U.S. models, charging as little as USD 0.18 per million tokens, compared to an average of USD 4 for top models, the Citi note showed. BlueRock's Byun noted open source models, once over a year behind, are now roughly four months adrift, with the gap expected to narrow.

Still, some analysts believe security concerns surrounding Chinese models will likely impede widespread enterprise adoption, especially in sensitive areas like cybersecurity. So businesses are expected to adopt a cloud computing strategy, diversifying across multiple providers for optimal fit and price.

Val Bercovici, chief AI officer at WEKA, stated that open source models are "90 per cent as good at 10 per cent of the price". He added that premium tokens are unnecessary for every level of effort.

Rising AI costs prompt businesses to seek cheaper alternative models.
Usage based pricing causes unpredictable and often higher bills for organisations.
Open source and Chinese AI models are gaining traction due to lower costs and closing capability gaps.
OpenAI and Anthropic are expected to engage in a price war amidst IPO preparations.
Security concerns for certain models may influence enterprise adoption, leading to diversified provider strategies.

Source: Reuters