The head of energy at Google's data centers stated this week that the cost of space data centers could potentially be up to 100 times that of ground-based data centers. Google itself is testing AI server chips in orbit, and this statement has poured a bucket of cold water on the hot concept of space computing power.
"Before embarking on space exploration, we should first thoroughly explore all feasible options on Earth," said Amanda Peterson Corrio at the WTF summit. This summit brought together senior female professionals from the technology and finance industries. At the time she expressed this view, the orbital data center planning of SpaceX was one of the key factors supporting its market value of approximately 1.8 trillion dollars.
Petersen Corrio also mentioned that the growing opposition to on-premises data centers in American society can be largely attributed to a group of newer companies with less experience, which should be subject to higher regulatory standards. She stated that before this wave of AI computing power hype, companies that had been deeply involved in data center construction for many years "operated in a more standardized manner," and related projects rarely sparked controversy.
The launch cost of putting computing hardware into orbit is “extremely high, up to 100 times the cost of ground construction,” said Peterson Corrio. In addition to costs, operators also need to address the challenges of maintaining and upgrading equipment in orbit. On the ground, data center staff can replace servers at any time and rewire and debug the system.
Despite a cautious attitude towards space computing power, Google's research team officially announced this month that it will collaborate with satellite imagery company PlanetLabs to explore the feasibility of large-scale space data center clusters using satellites. Under a project called “Suncatcher (Light Harvesting Plan),” Google has already launched a small prototype space server into space, which is equipped with AI tensor processing units (TPUs) custom-designed in collaboration with Broadcom.
Last autumn, Google researchers involved in the Lightcatching Project, along with the company's chief researcher James Manica, published a paper stating that if the launch cost drops to $200 per kilogram or less, the computational power cost corresponding to that level of power consumption would be expected to be roughly comparable to that on Earth. Google researchers believe that this threshold could be achieved around 2035, although the exact timing depends on the research and development progress of the SpaceX starship and the speed at which launch frequencies increase.
However, other companies are more optimistic. Elon Musk, the CEO of SpaceX, stated that the company could build a space data center as soon as next year; Jeff Bezos, the founder of Amazon and the leader of Blue Origin, a SpaceX competitor, predicts that space data centers with a power consumption of 1 gigawatt will emerge within ten years. There are also reports that Google and SpaceX have discussed plans to collaborate on building in-orbit data centers.
Peterson Corrio did not completely reject the concept of space-based computing power; she described this direction as “very interesting.” The reason Google continues to invest in such high-risk, cutting-edge projects is that the company is accustomed to “thinking about what will happen 10, 20, 30 years from now.” However, the implementation of this technology “is by no means imminent.” Sam Altman, CEO of OpenAI, was even more direct in his statement, believing that moving data centers into space is “quite unrealistic” under current circumstances, and he questioned Musk’s aggressive timeline.
The real bottleneck: approval and licensing, not electricity
In response to the views expressed by Musk and others that the US power supply is not sufficient to support the expansion of AI's computing power, Peterson Corrio has put forward a different perspective.
"Is there enough electricity? Of course," she said. The real bottleneck lies in the approval processes and lengthy evaluation studies required for the project to be connected to the grid. Laboratory X under Alphabet has launched the Tapestry project, and is collaborating with PJM, the largest power grid operator on the East Coast of the United States, to use AI to accelerate the relevant evaluation processes.
In Texas, the state's electric power regulatory agency has slowed down the approval process for some data center projects and strengthened supervision to prevent new computing facilities from increasing the burden on the power grid and driving up electricity prices.
Other senior executives in the industry also pointed out that the utilization rate of the power grid itself is far from reaching an ideal level. Brian Janos, co-founder of Cloverleaf Infrastructure, a data center land and power service provider, mentioned in a recent interview that on average, about 45% of the power grid's capacity is idle for long periods of time, which is to cope with peak electricity consumption. For the vast majority of the day, the power grid has excess capacity to meet the electricity demands of data centers.
It is worth mentioning that in terms of the energy utilization efficiency ( PUE ) indicator, Google has an advantage over its competitors in the cloud business sector. This indicator measures the efficiency of data center computing power, representing the ability to extract more computing power from the same amount of electricity.
Regarding the wave of anti-data center sentiment that has emerged within the United States, Peterson Corrio attributes it to "a large number of new players entering the market, who lack industry experience and mature practices when it comes to acquiring land and building computing power projects," but she did not name any specific companies.
In recent years, a large number of novice developers have flocked to the AI data center market. However, it's not just new players; even projects from well-known companies like Oracle have encountered resistance from local residents.
Higher industry standards
In any case, she supports the industry in introducing regulatory rules and setting higher standards. She did not propose specific provisions, but gave an example of Google increasing its investment in the local communities where its projects are located, while also putting in place contingency plans to avoid the risks associated with unfinished projects.
This includes a commitment to continuously bear the costs of grid upgrades over a period of 15 to 20 years, in order to prevent local communities from being 'forced to pay' after the project is halted. This cooperative model, which involves making such upfront commitments to cover the costs of grid infrastructure, has already been implemented in more than nine states in the United States and is 'expanding at an extremely fast rate.'
The supporting standards also include: Google paying the full amount of electricity bills and property taxes; paying a deposit, which will be used for land restoration if the project is abandoned; and canceling the confidentiality agreements signed with local governments.
The context of her statement this time is that investors are re-evaluating the valuation and viability of a group of emerging AI cloud service providers: most of these companies have received substantial support from NVIDIA and have embarked on the construction of data centers on a large scale, borrowing heavily without having a long-term operational track record.
The companies that Peterson Corrio mentioned need to have stricter standards most likely include this type of manufacturers. At the same time, Google has also revealed that its own computing power requirements will increasingly rely on third-party computing resources, but it has not disclosed the specific partners. SpaceX is one of them. Google needs external computing power because the speed at which it builds its own data centers cannot keep up with demand.
Responsible Editor: Jiang Xuesi
VIP Course Recommendations
APP Exclusive Live Streaming
Popular Recommendations
Hide the 24-hour rolling news updates and videos on finance and economics. For more benefits for fans, scan the QR code to follow ( sinafinance ).
