Broadcom chief questions enterprise AI returns as financing needs grow

Broadcom chief executive Hock Tan has questioned whether businesses’ current use of generative artificial intelligence is delivering sufficient returns, while warning that financing could become the industry’s next constraint.

The Business Times reported his comments at DBS’s Asia Future Forward Conference in Singapore on October 12. Hock Tan and Intel chief executive Lip-Bu Tan took part in a panel, moderated by DBS chief executive Tan Su Shan, on where the next trillion dollars of AI investment would come from.

Hock Tan said many companies had adopted AI in some form but were still looking for applications with a worthwhile return on investment. Current generative AI uses were often subtle and did not necessarily improve productivity, he said. Generative systems produce content, while productivity concerns what a business can produce with its resources.

Lip-Bu Tan said investors were increasingly seeking tangible impact and stronger returns, given the scale of money entering the sector. The discussion linked investment in the technology to the economic benefit companies can obtain from using it.

Hock Tan argued that the present boom was concentrated in hardware providers. Future value, he said, could lie with developers of frontier models—highly capable general-purpose systems—which might displace existing software businesses. In his view, supplying the hardware alone would not sustain the expansion.

That shift would also raise a financing question. Model developers could turn to the same sources of capital already funding infrastructure providers, he said, making access to money a possible bottleneck.

Lip-Bu Tan remained optimistic about the technology’s potential, predicting an impact larger than the internet’s and greater efficiency from software agents as more data become available. Neither chipmaker leader said he saw signs of a slowdown, despite fears of an AI bubble.

Hock Tan also addressed investors’ disappointment with OpenAI’s latest revenue forecast. He described US$50 billion in annualised revenue as substantial for a startup and said the business model was viable for now. Annualised revenue expresses a revenue pace as a full-year amount, rather than necessarily recording a completed year’s sales.

He argued that some investors did not understand the business model and had set extreme expectations. Model companies were currently losing money because of heavy spending on training, he said, but greater scale could change that position.

Tan Su Shan offered DBS’s account of a different relationship between use and cost. The bank had quintupled its AI consumption while reducing overall token costs, she said. Tokens are units such as words, word fragments or punctuation processed by a model; total usage and the price paid for it can move differently.

She argued that successful enterprises would be those adopting the technology to improve return on investment and return on equity. The first compares the benefit of an investment with its cost; the second relates profit to shareholders’ capital. More usage is therefore a different measure from either business return.

In a separate conference panel on capital flows, Franklin Templeton chief executive Jenny Johnson emphasised the transformation required across industries. Her focus was on the businesses AI could enable, rather than only the suppliers of computing hardware.