When will AI investments yield returns? JPMorgan Chase is optimistic about Google and Amazon
TL;DR
· Google Cloud's growth expectations have been significantly raised, primarily due to the inclusion of more TPU sales in the model, and cannot all be understood as an acceleration in traditional cloud service demand.
· JPMorgan expects AWS to grow by 40% in 2027, supported by factors including the expansion of its partnership with OpenAI, new GPU deployments, and AI computing pricing capabilities.
· AWS price increases are concentrated on reserved AI computing, which cannot be directly applied to all cloud services, nor does it mean that overall profit margins will increase proportionally.
· Meta's revenue expectations have been slightly revised upward, but profit forecasts have been downgraded, with expenses and capital expenditures still constraining return on investment.
· Fiverr is facing both a loss of search traffic and the impact of low-value services, with AI-generated revenue becoming differentiated across various platforms.
Editor's Note: Google Cloud's growth expectation for 2027 has been raised from 70% to 107%; AWS's growth expectation for the same period has been raised to 40%. Ahead of the third-quarter earnings report, JPMorgan has given a more positive assessment of both cloud businesses.
The factors supporting these adjustments are quite specific: more TPU sales have been included in the Google Cloud model, while AWS benefits from the expansion of AI collaborations, new GPU deployments, and rising prices for reserved computing. For cloud vendors, the basis for assessing AI demand is increasingly falling on revenue structure, delivery capability, and pricing levels.
However, the upward revision of growth expectations does not mean that investment returns have been fully validated. In the same report, Meta faced a profit downgrade due to legal expenses and higher spending expectations, while Fiverr continues to bear the impact of AI on traffic entry and low-value services. The differentiation of AI investments has begun to manifest in different aspects of revenue, profit, and cash flow.
The following is the original text compiled:
Google Cloud Growth Expectations Double, Key Increment from TPU Sales
Before the third-quarter earnings report, JPMorgan significantly raised its growth expectations for Google Cloud, with the core increment coming from the sales of its self-developed AI chip, TPU.
The bank raised its expectation for Google Cloud's year-on-year revenue growth in Q3 2026 from 88% to 100%, and in Q4 from 90% to 105%. A more significant adjustment occurred in 2027: the full-year growth expectation was raised from 70% to 107%, and from 55% to 63% for 2028. These figures are institutional forecasts and do not yet reflect the company's announced performance.
JPMorgan stated that the main reason for the upward revision was the inclusion of more Tensor Processing Unit (TPU) sales in the model and believes that the adjusted forecast may still be conservative. TPUs are specialized chips designed by Google for machine learning computations, and their sales growth is becoming an important variable in the bank's assessment of Google Cloud's revenue.
This also means that understanding Google Cloud's growth requires further observation of revenue composition. The 107% growth forecast includes more contributions from TPU sales and cannot all be attributed to an acceleration in traditional cloud service consumption. The profit margins, delivery pace, and cash recovery characteristics of different businesses will also affect how much revenue growth can ultimately be converted into profit.
At the group level, JPMorgan raised Alphabet's 2027 U.S. Generally Accepted Accounting Principles (GAAP EPS) forecast by about 3% to $14.05, and by about 4% to $17.29 for 2028; the target price for December 2027 was raised from $420 to $425, maintaining an "Overweight" rating, corresponding to about 25 times the expected EPS for 2028.
The significant increase in cloud business growth expectations contrasts with a relatively moderate upward revision in group profit forecasts. The extent to which growth in a single business can bring about profit elasticity still depends on its revenue scale, cost structure, and the performance of other group businesses.
Financing has also been included in this adjustment. JPMorgan incorporated the $5 billion debt financing completed by Waymo in early October into the model and assumed an additional $80 billion in debt financing in 2027. The latter is a predictive assumption and should not be regarded as announced or completed financing.
AWS Growth Logic Extends to Price Increases, New Computing Power Still Needs to Be Realized
For Amazon, JPMorgan expects AWS revenue to accelerate growth to 40% in 2027, and believes this forecast may still be conservative.
Its judgment is supported by two main factors: first, the expansion of the partnership with OpenAI, and the additional deployment of 2 million GPUs from 2027 to 2028; second, the pricing capability of AI computing power. New computing power determines potential delivery scale, while price affects the revenue that can be generated from the same resources.
According to the report, AWS raised the price of reserved GPU computing power by 15% starting October 7, marking the fourth consecutive quarter of price increases for related AI reserved computing. The report also noted that prices for H100/H200 and B200/B300 have cumulatively increased by 40% to 75% since the beginning of the year.
Reserved computing refers to service arrangements where customers pre-book a certain computing capacity. The price increases mentioned in the report are concentrated on specific GPUs and reserved AI computing power, and cannot be used to determine that all AWS products are being increased in price by the same magnitude.
JPMorgan also emphasized that customer demand has a certain breadth: AWS's computing power distribution is relatively decentralized, with the largest customer accounting for only a single-digit percentage. Therefore, its growth argument includes both increments brought by large AI collaborations and demand support from a broader customer base.
These factors collectively drove the bank to raise its forecasts for Amazon's 2027 and 2028: group sales are expected to increase by about 1% to 2%, operating profit by about 3% to 5%, and GAAP EPS by about 2% to 3%. JPMorgan maintains a target price of $365 for December 2027 and an "Overweight" rating, based on a valuation of about 27.5 times the expected EPS of $13.34 for 2028.
The revenue and profit revisions support its positive stance, but AWS still needs to translate collaborative expansions and deployment plans into actual deliveries. The report also lists the early stage of monetizing generative AI, limited visibility, and the potential impact of ongoing investments on profit margins as risks. Price increases provide revenue support, but their ultimate profit contribution still depends on equipment, operational, and delivery costs.
Meta's Revenue Slightly Increases, Expense and Cash Flow Pressure More Prominent
Compared to the upward revisions for the two cloud businesses, Meta's adjustment mainly reflects increased expenses.
JPMorgan incorporated a $10 billion legal expense provision into its Q3 2026 model. The report attributes this to a settlement reached by Meta with 52 state attorneys general and notes that this expense was not included in the company's guidance of $165 billion to $169 billion for total expenses provided in the second-quarter earnings report.
As a result, the bank lowered its GAAP EPS forecast for Meta in 2026 by about 10%, and by about 2% for 2027, to $27.99 and $31.08, respectively.
At the same time, JPMorgan raised its total GAAP expense forecast for 2027 to about $210 billion, an increase of about 18% year-on-year; if the aforementioned $10 billion legal expense is excluded from the 2026 base, the expense increase for 2027 is about 25%. The lower apparent growth rate is partly influenced by the elevated base from one-time expenses in the previous year.
On the revenue side, there was a slight improvement. The bank raised its revenue forecast for Meta in 2027 by about 0.9% to $307.9 billion, but this increment is not enough to fully offset the impact of the expense forecast revision.
Capital expenditures and cash flow provide another layer of observation. According to the report model, Meta's capital expenditures for 2027 are about $242.7 billion, with free cash flow (FCF) at about negative $72.8 billion. Negative free cash flow indicates that operating cash is expected to be insufficient to cover capital expenditures during the same period.
Meta's core contradiction lies in the need for improvements in advertising and AI monetization to catch up with the massive scale of investment.
JPMorgan believes that improvements in AI advertising ranking and recommendation capabilities are enhancing user engagement and advertising returns, and is optimistic about the monetization path for consumer and enterprise intelligent agents and model interface services, thus maintaining an "Overweight" rating and a target price of $920.
However, the bank also clearly warns that if AI monetization lags, ongoing investments will suppress operating profit, EPS, and free cash flow. Maintaining the rating reflects confidence in longer-term returns, but recent spending pressures still exist.
Fiverr Faces Another Impact, AI Revenue Will Not Be Distributed Equally
The downgrade for Fiverr in the report stands in stark contrast to cloud vendors: while cloud vendors are trying to capture AI computing demand, service transaction platforms need to cope with the impact of AI on traffic and transactions.
JPMorgan cited SimilarWeb data indicating that Fiverr's U.S. website traffic decline year-on-year expanded from 10% in Q2 to 22% during the period for which data was obtained in Q3. The monthly data listed in the report shows a decline of 19% in July and 24% in August, so this statistic cannot be directly taken as the complete result for Q3.
In JPMorgan's view, changes such as AI summaries in Google Search are reducing opportunities for users to click on paid or organic search links. Even if Fiverr invests in Generative Engine Optimization (GEO) to try to increase exposure in AI answers, the click-through rate from large language models remains lower than traditional search, thus putting pressure on the platform's ability to acquire potential customers.
The service supply side is also facing adjustments. Fiverr is transitioning from a transaction market primarily focused on low-value projects to a higher-end, higher-value work platform. However, the report states that high-value business currently accounts for only 15% of the platform's gross merchandise volume (GMV), and management expects the transformation to take at least six quarters.
Based on this, JPMorgan expects Fiverr's revenue in 2027 to decline by about 15%, the same as in 2026, and has lowered its revenue forecasts for 2027 and 2028 by about 6%, with adjusted EBITDA forecasts lowered by 21% and 14%, respectively. The bank maintains a "Neutral" rating, with the target price reduced from $11 to $10.
The report also adjusted the forecasts for the ride-hailing platform. Uber's restructuring involves about 3,300 people, and JPMorgan estimates annual savings of over $600 million, but considering the company's plan to reinvest funds into price reductions, service options, and growth businesses, it has not yet counted this as a net cost saving. Lyft, on the other hand, faces additional costs and cash expenditures due to disputes over driver classification; the bank maintains a neutral rating and a target price of $19.
These adjustments collectively illustrate that assessing the returns of AI and platform businesses requires a detailed examination of revenue sources, cost changes, and cash flows. Google Cloud's TPU sales, AWS's computing power delivery and pricing, Meta's expenses and capital expenditures, and Fiverr's traffic and transformation progress will each determine whether their respective growth logic can be realized.