Free tools Windows power users keep installed
One-click scans. No signup required.
The AI long game is not just a contest to build the most capable model. It is a fight to control the infrastructure, distribution and customer relationships that can turn AI use into durable revenue. In a July 12, 2025, GeekWire discussion, Baird senior research analyst Colin Sebastian used that lens to compare Amazon, Google and Meta. Their later disclosures show how much they are committing—but spending and product launches alone do not prove that the investments will pay off.
What the AI long game means
Sebastian’s framework is about advantages that can compound over years: engineering talent, computing capacity, proprietary data, broad distribution and the financial ability to invest before the payoff is clear. The strategic prize is not only answering questions. It is becoming the platform people use to search, shop, work and complete tasks—and doing so at an attractive cost.
That differs from short-lived excitement over a benchmark, viral demo or model launch. A durable advantage would show up in repeat usage, customer retention, lower costs per task, new or protected revenue, and returns sufficient to justify infrastructure spending. A company can lead on one dimension and still lose on another: a strong model may lack distribution, while a widely used product may fail to monetize.
Sebastian discussed the framework on GeekWire’s July 12, 2025 episode. The episode is an analyst-led strategic discussion, not a stock rating or a guarantee about which company will win.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →#1 Best Overall
Amazon: infrastructure, commerce and Alexa
AWS can sell the tools behind AI
Amazon’s AI strategy extends well beyond its consumer assistant. AWS can earn revenue from compute, model training and inference, model hosting, developer services and enterprise deployment. Bedrock and Amazon’s custom chips, including Trainium and Graviton, fit the broader effort to offer customers infrastructure and tools while managing the cost of running workloads.
Amazon’s 2025 Form 10-K linked rising technology and infrastructure spending with AI and machine learning. CEO Andy Jassy’s 2025 shareholder letter said Amazon expected roughly $200 billion in total capital expenditure in 2026, with much of the AWS-related investment expected to be monetized in 2027–2028. That is a company outlook for total capex, not a claim that the entire amount is AI spending; the timing and returns remain uncertain.
Alexa could become a transaction interface
In the 2025 discussion, Alexa+ represented a possible shift from voice commands and answers toward an assistant that can take actions. Jassy described that ambition in Amazon’s account of its generative-AI strategy. If a reliable assistant can help users discover products and complete tasks, it could reinforce Amazon’s marketplace, Prime relationship and device ecosystem.
The difficult test is whether Alexa can make transactions dependable and useful enough that people choose it—and whether those transactions generate incremental, profitable business. Amazon also faces the possibility that shoppers will begin with a general-purpose assistant instead of Amazon’s own interface. In that case, Amazon could still benefit by providing the goods or infrastructure, but lose influence over discovery and the customer relationship.
Amazon’s trade-off
AWS gives Amazon a route to monetize AI workloads even if a particular consumer assistant does not become dominant. The counterweight is capital intensity: data centers, chips and related infrastructure bring operating costs and depreciation, and capacity built ahead of demand may not earn an attractive return. Alexa’s commercial value, the cost of competing in models and the extent to which AI boosts AWS revenue rather than expense are all questions the long game must answer.
Rank #2
Google: reinventing search without breaking its economics
AI is both a search threat and a search opportunity
Google’s central tension is that generative answers can make search more useful while changing the results page that has supported its advertising business. In the GeekWire episode, Sebastian argued that Google could face losses in search share or query volume in some markets as it responds with AI Overviews, AI Mode and Gemini. That was his assessment, not proof of a universal or permanent decline.
Conversational answers may satisfy a user without a traditional click. That can affect publisher referrals, ad placement and commercial intent; generating an answer may also require more computing than serving conventional results. Alphabet’s 2025 Form 10-K recognizes that AI products could change monetization, revenue growth and margins, and that serving AI requires more compute, energy, equipment and network capacity. The filing also points to increased technical-infrastructure investment in 2026.
Integration is an advantage, not a guarantee
Google can place AI across search, Android, Chrome, YouTube, Maps, Gmail, Workspace and Google Cloud, alongside its models and infrastructure. That reach gives it many ways to put AI in front of users and sell services to organizations. Google Cloud is a potential counterweight if consumer search economics change: Alphabet’s filing identifies enterprise AI services as part of its expanding cloud offering.
Recommended Free Tools
But broad distribution does not settle the hardest question. Google must move users toward more capable AI interfaces while retaining trust and finding a sustainable way to monetize them. If AI increases engagement but reduces valuable clicks or raises serving costs faster than revenue, usage growth alone would not validate the strategy.
Meta: an assistant within the social graph
In Sebastian’s account, Meta’s distinctive bet was to combine AI talent with its vast consumer surfaces. Meta AI across Facebook, Instagram, WhatsApp and Messenger could turn an existing social and messaging relationship into an assistant people use throughout the day. AI can also support recommendations, advertising optimization and tools for creating images, video and other content.
That strategy is different from AWS’s role as an infrastructure seller or Google’s need to defend search. Meta’s opportunity is to make AI improve its social products and potentially become a new interface for questions, recommendations and tasks. Messaging could also provide a setting for agents, while better recommendations and creative tools may strengthen engagement and advertising.
The open question is the business model. Meta AI might become a valuable feature that keeps people engaged, a large assistant platform, a gateway to commerce and ads, or a costly research and infrastructure effort with limited direct revenue. Talent recruitment and model capability are inputs; durable usage and measurable economic returns are the test.
Compare control points, not just model rankings
The following is a strategic comparison, not a direct company ranking. It summarizes the companies’ established platforms and the business models most exposed to their AI plans.
| Control point | Amazon | Meta | |
|---|---|---|---|
| Consumer distribution | Marketplace, Prime, Alexa and devices | Search, Android, Chrome and YouTube | Facebook, Instagram, WhatsApp and Messenger |
| Enterprise distribution | AWS, Bedrock and developer services | Google Cloud and Workspace | Less central to its disclosed consumer-platform strategy |
| AI interface ambition | Alexa and shopping or task agents | AI-assisted search, Gemini and Workspace | Meta AI within social and messaging products |
| Potential data advantage | Commerce, logistics and cloud usage | Search, video, maps and productivity activity | Social content and interactions |
| Established monetization at stake | AWS, commerce and advertising | Search advertising, Cloud and subscriptions | Advertising, with possible commerce and subscription opportunities |
| Central strategic risk | Capex burden or loss of control over shopping discovery | Search cannibalization and higher AI-serving costs | High talent and infrastructure costs without clear assistant economics |
These are potential advantages, not proof that any company can use its data for every AI purpose or that an existing interface will retain users. Model quality matters, but it is only one part of a system that also needs affordable inference, trust, distribution and a way to capture value.
Why commerce discovery is an important test
Sebastian cited a Baird survey of Gen Z shopping discovery in which ChatGPT reportedly ranked first in the later survey, after TikTok had ranked first a year earlier; Amazon and Google had ranked first in earlier years. The episode summary does not provide the survey’s sample size, geography, question wording or methodology, so the result is best treated as a directional illustration—not evidence that ChatGPT has displaced either retailer or search engine in overall shopping.
Discovery is also not the same as a completed purchase. For an assistant to control commerce, it must handle practical details such as current prices, stock, shipping, returns and trustworthy recommendations. If users increasingly begin product research with an assistant, however, the first point of contact could move upstream from a retailer’s own site. That possibility matters to Amazon’s interface ambitions and to Google’s search business, even before it produces a clear winner.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →What the enterprise spending survey does—and does not—show
Sebastian reported that a Baird survey of 100 corporations found 87% planned to increase generative-AI spending over the following year, and no respondent planned to spend less. The accessible episode summary does not state the survey date, respondent profile, question wording or margin of error. The figures therefore describe the reported intentions of that sample, not all businesses or realized spending across the economy.
Planned budgets can indicate interest, but they do not establish that employees adopted the tools, that pilots reached production, or that customers and vendors earned a return. The more consequential signal is whether spending becomes repeat use and measurable value: lower costs, higher output, new sales or software and cloud revenue that exceeds the resources required to deliver it.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How this AI boom compares with the dot-com era
Sebastian’s comparison emphasized that today’s largest AI spenders are established companies with substantial businesses and cash-generating operations, unlike many unprofitable dot-com ventures. Their existing cloud, advertising, software and commerce products can also put AI into revenue-producing channels more quickly than a startup with no customers or distribution.
The analogy still has limits. Both eras feature large expectations, infrastructure bottlenecks, rapid investment and uncertainty about which new business models will endure. Financial strength at the parent-company level does not show that a specific data center, model or consumer product will earn an adequate return. For today’s giants, the risk may be less about immediate insolvency and more about overbuilding, weak monetization or spending defensively to avoid falling behind.
Best Value
What could confirm—or weaken—the long-game thesis
Investors and business leaders can evaluate the strategy through a small set of operating questions rather than relying on launch announcements or model rankings.
- Revenue and adoption: Are AI services generating recurring revenue, sustained cloud consumption or repeat consumer use? Do users complete tasks, or mainly try a feature once?
- Unit economics: Are inference costs falling relative to revenue per user, workflow or query? Do margins improve as use scales, after accounting for energy, equipment and depreciation?
- Incrementality: Does AI create new sales, cloud workloads or advertising value, or mainly shift activity from an existing product whose economics were already strong?
- Infrastructure utilization: Is customer demand supporting new capacity, and does cash flow recover after the capital-spending surge? Amazon’s stated expectation that much AWS-related 2026 investment would be monetized in 2027–2028 makes subsequent utilization and returns especially relevant.
- Interface control: Do people return to Alexa, Gemini or Meta AI to search, shop and complete tasks, or do third-party assistants become the default gateway?
- Reliability and trust: Can systems use tools accurately, protect privacy and recover from mistakes? Errors become more consequential when an assistant acts rather than merely generates text.
- Enterprise conversion: Do pilots become production deployments with identifiable productivity or revenue gains, rather than recurring experiments without a clear business case?
The thesis weakens if infrastructure arrives ahead of demand, model capabilities become interchangeable, enterprise pilots fail to scale, or AI answers cannibalize profitable activity without replacement revenue. Privacy limits, copyright disputes, regulation and reliability failures could also constrain data use or reduce user trust. Conversely, a company might lose a model comparison yet still win through distribution, or grow AI revenue while margins deteriorate; neither headline alone resolves the economics.
The key question is who captures value
Amazon, Google and Meta are pursuing different positions in the same evolving system: AWS sells capacity and tools, Google is rebuilding search and extending AI into cloud and productivity, and Meta is trying to make AI part of social and messaging experiences. Sebastian’s long-game lens directs attention away from a single model leaderboard and toward whether those positions create recurring, profitable relationships with users and customers.
The decisive evidence will be sustained adoption and returns—not the size of a spending plan or the promise of an assistant. Each company must show that AI strengthens its existing platform or creates a valuable new one without making its core economics worse.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
