Google’s chief scientist claims that AI is being unfairly blamed for the rise in his company’s carbon dioxide emissions, and says the tech giant’s efforts to switch to entirely clean energy by 2030 remains on track.
Earlier this month, the Mountain View-based biz published its annual environmental report, containing the admission that its CO2 emissions for 2023 were up 13 percent year-on-year, and up by 48 percent compared to the 2019 target base year.
The cause of this was identified as increases in datacenter energy consumption and supply chain emissions, which inevitably meant people started pointing fingers at AI, especially as Google is one of the internet companies that continue to ramp AI efforts over the past couple of years.
In an interview with Fortune during its Brainstorm Tech conference this week, Google’s Jeff Dean claimed that AI is not as responsible for rising datacenter carbon emissions as critics make claim.
“There’s been a lot of focus on the increasing energy usage of AI, and from a very small base that usage is definitely increasing,” Dean said. “I think people often conflate that with overall datacenter usage – of which AI is a very small portion right now but growing fast – and then attribute the growth rate of AI based computing to the overall datacenter usage.”
In other words, AI may be a rapidly expanding workload, but it currently represents only a small portion of the overall global datacenter burden, according to Google’s chief scientist.
He claimed Google was not giving up its commitment to be powered entirely by clean energy by the end of 2030, and said progress is “not necessarily a linear thing.” This is because some of Google’s efforts to source clean energy from providers will not bear fruit for several years.
The Chocolate Factory’s environmental report [PDF] states that its Scope 2 emissions, which include those from energy suppliers, increased 37 percent to represent 24 percent of its total carbon footprint for 2023.
However, if Google is matching 100 percent of its electricity consumption with renewable energy, why do its Scope 2 emissions show an increase instead of being zero? Google asked this question itself and says it is because its approach to renewable energy matching differs from the way Scope 2 emissions are calculated in the Greenhouse Gas (GHG) Protocol, among other reasons.
Google’s report also said that while the total electricity load across all its datacenters increased by roughly 3.5 TWh (17 percent) during 2023, the cloud and search biz maintained a global average of 64 percent carbon-free energy usage. Which doesn’t explain how it intends to get to 100 percent if its consumption keeps rising as quickly as it can add renewable energy sources.
And while Dean says that AI currently makes up only a small part of overall datacenter workloads, a recent report by Goldman Sachs estimated that AI will drive global datacenter power demand up by 160 percent come 2030, and that carbon dioxide emissions may more than double compared with 2022 levels.
Demand for higher performing (and therefore higher power consuming) AI servers continues apace, according to the latest report from TrendForce.
The market watcher estimates that AI server shipments will jump by 41.5 percent this year to exceed $187 billion, accounting for 65 percent of the total server market value.
As The Register has previously reported, this is because buyers such as the hyperscalers are purchasing fewer datacenter servers, but the ones they are choosing are costly, highly configured units bristling with GPUs.
Google isn’t the only company struggling with net-zero commitments. Microsoft admitted in its own Environmental Sustainability Report that its CO2 emissions are up 29.1 percent from the 2020 baseline, blaming the construction and outfitting of more datacenters to meet customer demand for cloud services. ®