Key Takeaways:
Barclays projects Google will spend $500 billion on capital expenditures by 2028, driven by the massive compute demands of training its Gemini 4 AI model.
Barclays raised its capital expenditure forecast for Google to $350 billion in 2027 and $500 billion in 2028, projecting two consecutive years of negative free cash flow as the company pours resources into training its Gemini 4 large language model.
"Google is paying a premium to lease external GPU capacity from Colossus at roughly $10 billion a month, effectively freeing its own first-party compute for Gemini 4 training," analysts at Barclays wrote in a July 22 report, raising their price target on Alphabet to $425 from $405 with an overweight rating.
The revised estimates follow Alphabet's second-quarter earnings, which beat Wall Street expectations on both the top and bottom lines. Revenue rose 24 percent year-over-year to $119.8 billion, while earnings per share reached $9.11, well above the $2.90 consensus. Google Cloud revenue surged 82 percent to $24.77 billion, and the unit's backlog jumped to $514 billion from $464 billion in the prior quarter. Yet the company's free cash flow turned negative for the first time in decades at negative $5.9 billion, according to AlphaSense data.
The spending trajectory puts Google on a path where capital expenditures will exceed operating cash flow by $65 billion in 2027 and $1.35 trillion in 2028, according to Barclays' model. Chief Financial Officer Anat Ashkenazi told analysts on the earnings call that AI infrastructure spending would continue to pressure both profit margins and cash flow, with capital expenditures expected to rise "significantly" in 2027. The question for investors is whether the cloud revenue acceleration — up 82 percent in the quarter — can eventually catch up to the spending.
The centerpiece of Barclays' revised outlook is a GPU leasing agreement with Colossus that costs Google roughly $10 billion per month. Barclays described the premium Google is paying as "eye-catching" but justified given the urgency of the compute shortage. The arrangement is structured as a temporary bridge: external GPUs serve Google Cloud Platform customers while Google's own tensor processing units are redirected toward training Gemini 4.
The scale of the training effort is unprecedented. Gemini API calls have reached 220 billion tokens per minute, up from 160 billion a quarter earlier and representing threefold to fourfold growth year-over-year. Barclays estimates Gemini now holds the No. 2 position in the large language model market alongside Anthropic, trailing only OpenAI.
Google's own capital expenditure guidance for full-year 2026 stands at $195 billion to $205 billion, up from a prior range of $180 billion to $190 billion. The company spent $44.92 billion on capex in the second quarter alone, slightly above the $44.66 billion analysts expected. Tesla also reported negative free cash flow for the first time in more than two years on Wednesday, underscoring the industrywide pattern of AI infrastructure spending outpacing cash generation.
Barclays characterized Alphabet as "the most AI-advanced company globally today," citing its vertical integration across application, model, infrastructure, and silicon layers. Google operates 13 products with more than 1 billion monthly active users, competes with Anthropic and OpenAI in models, faces Amazon Web Services and Microsoft Azure in cloud, and designs its own TPU chips.
This breadth creates a competitive moat that Barclays believes will become more apparent as cloud revenue accelerates in coming quarters. The bank raised its 2027 revenue and operating profit estimates by 5 percent each, reflecting higher cloud revenue and margins. However, the Colossus lease will compress cloud margins in the fourth quarter of 2026 and the first quarter of 2027 before the benefit of the capacity buildout flows through.
Alphabet shares, trading at roughly 24 times forward earnings, have gained about 10 percent this year but remain 15 percent below their May peak. Barclays' $425 target implies roughly 24 percent upside from current levels. The bank's valuation blends 25 times 2027-2028 expected earnings per share with 15 times EBITDA.
For investors, the tension is straightforward: Google is spending like a company that believes AI will generate returns far above its cost of capital, but the cash flow statement will remain deeply negative for at least two years. The bull case rests on cloud revenue growth sustaining its 80 percent-plus trajectory. The bear case is that the $500 billion 2028 capex figure keeps climbing as model sizes expand faster than inference revenue materializes.
This article is for informational purposes only and does not constitute investment advice.
