5 Companies That Could Benefit from the AI Boom in 2026: GPW and Wall Street

The boom in artificial intelligence is not limited to chip manufacturers. Money is also flowing to companies providing networks, power, and cooling for data centers, cloud computing, and ready-made applications for businesses. Therefore, the list of AI companies in 2026 should include various parts of this chain.

We analyze four companies from Wall Street and one conditional pick from GPW -- Text S.A.

This is a watchlist, not a ranking of expected returns.

• Nvidia remains the most direct exposure to AI infrastructure among this group, but a market capitalization of around $5.05 trillion implies very high market expectations. There are also conditional guarantees related to infrastructure for OpenAI.
• Broadcom profits from chips designed for specific customers and networks for data centers. Dynamic growth goes hand in hand with a high concentration of customers.
• Vertiv sells power and cooling systems needed for data centers, including the increasingly energy-intensive AI infrastructure.
• Microsoft monetizes AI at several levels: from Azure to Microsoft 365 Copilot. The scale of revenue is real, as are the increasing expenditures and pressure on cash flows.
• Text S.A. is a Polish, more conditional exposure to the application layer of AI. KPIs for the first quarter of the fiscal year 2026/2027 have improved, but the entire previous year saw a decline in revenue and profit.

Table of Contents:

• Why is the AI boom still driving stock markets?
• How did we select AI companies for 2026?
• Nvidia: central exposure to AI infrastructure
• Broadcom: custom chips and networks for data centers
• Vertiv: without power and cooling, AI doesn't work
• Microsoft: AI from data centers to applications
• Text S.A.: Polish AI company or just a candidate?
• Not just Nvidia. Who is really profiting from AI development?
• AI companies from GPW: does Poland have more beneficiaries?
• The biggest risks of investing in AI companies
• Is it too late to invest in AI by 2026? {#inside-nav}

Why is the AI boom still driving stock markets? {#1}

The most important fuel for the AI market today is not declarations but the spending of the largest tech companies.

According to the International Energy Agency, the five largest tech companies allocated over $400 billion for capital expenditures in 2025, and this amount could increase by about 75% in 2026.

This is an estimate from the IEA, not a sum reported under one identical definition.

The scale is also evident in reports covering the quarter ending June 30, 2026. Amazon, Alphabet, Microsoft, and Meta spent a total of about $165.0 billion in cash on fixed assets during this period.

The individual amounts were $54.2 billion, $44.9 billion, $35.8 billion, and $30.1 billion, respectively.

These are not pure AI expenditures: Amazon also finances logistics, and the reporting definitions of companies are not fully identical. However, the direction remains clear.

Data sources for the chart: reports from Amazon, Alphabet for the first half and first quarter, Microsoft and Meta for the first half and first quarter. For Alphabet and Meta, the value for the second quarter was calculated as purchases for the half-year minus the first quarter.*

Expenditures create demand on several layers simultaneously:

• accelerators and complete computing systems,
• switches, optics, and networks connecting thousands of chips,
• power, cooling, and physical infrastructure for data centers,
• cloud, models, and tools for developers,
• applications that users or businesses pay for.

The second signal is monetization.

The annual revenue of the entire Azure has exceeded $100 billion and increased by 41%, but Microsoft does not separate the sales generated by AI in this figure.

More than 30 million paid Microsoft 365 Copilot licenses confirm the application monetization, although the company did not disclose its value.

At the same time, the bill is getting bigger.

Global energy consumption by data centers is expected to rise from 485 TWh in 2025 to about 950 TWh in 2030, according to the central scenario of the IEA.

Building data centers requires energy, connections, transformers, permits, and specialized workers.

For investors, the key question is not "will AI be bigger?" but who can turn these investments into sustainable cash flow.

The practical side of end-user demand is described in the CrypS.pl guide on automating work in companies using AI.

How We Selected AI Companies in 2026

Each company had to meet most of the following conditions:

• show measurable revenue, demand, or AI-related KPIs,
• have an operating business and available financial data, not just a product announcement,
• occupy a significant place in the value chain,
• have a specific catalyst for the upcoming quarters,
• disclose risks that could undermine the thesis,
• be accessible to Polish investors through the GPW or the US market.

We selected one company listed on the GPW and four companies listed in the United States.

The second Polish candidate would lower the quality of the selection: interesting AI products can be found on the Warsaw Stock Exchange, but they usually lack scale, profitability, or separately measured monetization. We will return to this in a separate section.

Valuation also does not boil down to a single "magic" number.

The multiplier must be read together with the growth rate, quality of profit, and capital intensity. A good starting point is the CrypS.pl guide on how to value technology companies.

Nvidia: Central Exposure to AI Infrastructure

Nvidia no longer sells just individual graphics processors.

Its offering includes accelerators, complete server systems, NVLink connections, InfiniBand and Spectrum-X networks, and software needed to train and run models.

This is important because a customer building a data center does not just buy a chip.

They need a whole system that allows connecting thousands of processors and effectively utilizing their power. Nvidia's software ecosystem increases the cost of switching suppliers.

Nvidia's Q1 Results for Fiscal Year 2027

In the first quarter of fiscal year 2027, ending April 26, 2026, Nvidia achieved:

• $81.6 billion in revenue, up 85% year-over-year,
• $75.2 billion in data center revenue, up 92%,
• $60.4 billion from sales of computing solutions for data centers,
• $14.8 billion from data center networking, up 199%,
• 74.9% GAAP gross margin.

The company projected around $91 billion in revenue for the next quarter, with a margin of plus or minus 2%.

The forecast did not include revenue from sales of computing solutions for data centers in China.

Data is sourced from Nvidia's Q1 report for fiscal year 2027.

Catalysts, Valuation, and Risks

The catalyst remains the transition of customers to the next generations of systems and the growing demand for inference, meaning the daily running of ready-made models.

Nvidia is developing the Rubin platform, networks, and processors complementing GPUs, allowing it to increase the value of the entire set sold to the customer.

According to CrypS.pl calculations, the closing price of $208.48 on August 24 multiplied by about 24.2 billion shares gave a market capitalization of approximately $5.05 trillion.

Such scale means that the market assumes not only further growth but also the maintenance of high margins and technological advantages.

The most tangible counterargument is China.

Nvidia stated in its 10-Q report that restrictions have effectively closed its access to the Chinese advanced computing solutions market.

In the analyzed quarter, there were no shipments of Hopper Data Center products to China, while a year earlier they amounted to $4.6 billion.

Nvidia has also begun supporting infrastructure development with its own balance sheet.

In the August 8-K report, it revealed guarantees for the residual value related to the lease of approximately 4.25 GW of infrastructure, which is expected to be leased by OpenAI.

The total limit for conditional payments is $105 billion, but this is neither current debt nor planned CAPEX.

The obligation may be triggered upon OpenAI's insolvency or payment default and only in the event of a shortfall after re-leasing or selling assets.

OpenAI has committed to reimbursing Nvidia for the amounts actually paid.

Nvidia also announced a $1.5 billion investment in SB Energy and retained the option to support an additional approximately 3.8 GW.

This strengthens potential demand for its systems but increases counterparty risk, infrastructure utilization, and its residual value.

There are also competitors like AMD, proprietary chips from cloud companies, dependence on Asian production, and the risk that the current pace of expenditures may not be sustainable.

Among the analyzed companies, Nvidia has the most direct exposure to the AI boom, but the price and new conditional obligations leave little room for disappointment.

Broadcom: Custom Chips and Networks for Data Centers {#4}

Large cloud companies do not always want to buy only general-purpose processors.

At the right scale, it is profitable for them to design a chip tailored for a specific task.

Broadcom helps create such custom accelerators, known as ASICs.

The second pillar consists of networking solutions. Simply setting up multiple processors in a data center is not enough.

It is also necessary to transmit vast amounts of data between them with low latency.

Broadcom's Results for Q2 of Fiscal Year 2026

In Q2 of fiscal year 2026, Broadcom reported:

• $22.19 billion in revenue, up 48% year-over-year,
• $10.8 billion in AI semiconductor revenue, up 143%,
• $15.24 billion in adjusted EBITDA, corresponding to 69% of revenue,
• $10.26 billion in free cash flow, or 46% of revenue,
• $2.44 diluted non-GAAP EPS.

For Q3, management forecasted approximately $29.4 billion in revenue and $16 billion in AI semiconductor revenue.

Achievement would mean an increase in the latter category of over 200% year-over-year.

Catalysts, Valuation, and Risk

The most important catalyst is the next generations of chips designed for the largest cloud customers.

Broadcom also benefits from the increasing number of connections within AI clusters.

The more processors work together, the more important a fast network becomes.

According to CrypS.pl calculations, a price of $358.76 multiplied by approximately 4.76 billion shares gives Broadcom a market capitalization of approximately $1.71 trillion.

For additional reference: $358.76 divided by four times the quarterly diluted non-GAAP EPS of $2.44 gives about 36.8.

This is merely a mechanical multiplier of the current pace, not a full-year forecast.

The main risk is concentration.

According to Broadcom's 10-Q report, the five largest end customers accounted for about 45% of revenue in the first half of fiscal year 2026.

Separately, direct sales to one distributor customer represented 42% of total revenue, up from 29% a year earlier.

Losing a project, disrupting the distribution channel, or delaying implementation can significantly change results. A chip designed for one customer cannot always be sold to someone else.

Broadcom also had $62.7 billion in long-term debt. The profitable infrastructure software business generates cash, but in the last quarter, it grew significantly slower than semiconductors.

The company provides exposure to a segment of the market outside of universal GPUs, but investors are taking on concentration risk and very high expectations.

Vertiv: Without Power and Cooling, AI Doesn't Work {#5}

AI processors consume a lot of energy and generate heat.

The more power concentrated in a single server rack, the harder it is to power the equipment and maintain the appropriate temperature.

Vertiv provides uninterruptible power supply systems, power distribution, liquid cooling, and services needed for data centers.

The company does not need to predict which language model will win. It benefits from the sheer growth in density and scale of infrastructure.

Vertiv's Results and Forecast

In Q2 2026, Vertiv achieved:

• $3.27 billion in sales, up 24% year-over-year,
• 18% organic growth,
• 22.6% adjusted operating margin, up 4.1 percentage points,
• $925 million in adjusted free cash flow,
• $1.52 adjusted earnings per share, up 60%.

Management raised the full-year sales forecast to $13.8-14.2 billion and expected $6.65-6.75 adjusted EPS. Details can be found in Vertiv's Q2 2026 results.

Catalysts, Valuation, and Risks

Increasing power density enhances the value of infrastructure per rack. Transitioning to liquid cooling is particularly important, as traditional air cooling may not suffice for the most loaded systems. Vertiv can thus benefit from both the construction of new data centers and the modernization of existing facilities.

According to CrypS.pl calculations, the price of $254.97 on August 24 divided by $6.70, which is the midpoint of the adjusted EPS forecast for 2026, yielded about 38 times the projected earnings. This is a demanding multiple for a physical infrastructure provider.

Vertiv already reported transitional bottlenecks in the supply chain and delays in multi-stage projects in Q2. Risks also include order cancellations, tariffs, and increases in inventory and working capital. The order book is not revenue until the project is delivered and settled.

Vertiv shows that an AI company does not need to have its own model or processor. The thesis is based on the physics of data centers, but the valuation no longer resembles that of a traditional device manufacturer.

Microsoft: AI from Data Center to Applications {#6}

Microsoft occupies several floors of the value chain simultaneously. Azure sells computing power and tools for building applications. Microsoft 365 Copilot, GitHub Copilot, and Dynamics are trying to turn AI into a paid feature for businesses.

Distribution is an advantage here. Microsoft already has relationships, contracts, and business customer data, so it does not need to acquire users from scratch.

Microsoft’s Results for Fiscal Year 2026

In the fiscal year ending June 30, 2026, Microsoft reported:

• $331.84 billion in revenue, up 18%,
• $155.24 billion in operating income, up 21%,
• $214.4 billion in Microsoft Cloud revenue, up 27%,
• Azure revenue growth of 41% for the entire year,
• $66.99 billion in free cash flow according to CrypS.pl, calculated as $182.94 billion in operating cash flows minus $115.95 billion in cash purchases of fixed assets. This is not the FCF reported by Microsoft nor the full CAPEX of the company.

In Q4, Azure grew by 43%, and the number of paid Microsoft 365 Copilot seats exceeded 30 million.

Data comes from Microsoft's 10-K report and the Q4 earnings conference.

Microsoft also revealed in the 10-K report $24.1 billion in revenue from commercial agreements with OpenAI in the fiscal year 2026, including payments from the revenue-sharing mechanism, and $6.0 billion in receivables at the end of the year.

OpenAI is a related entity for Microsoft. The scale of this relationship enhances exposure to AI but simultaneously increases concentration and complicates the assessment of the quality and independence of part of the revenue.

This amount should not be equated with Azure revenues or added to them.

Catalysts, Valuation, and Risks

Microsoft earns both when a customer creates their own application in Azure and when they purchase a ready-made Copilot.

The catalyst will be the pace of transitioning pilots to full contracts and the use of agents in daily work.

Excluding the impact of investments in OpenAI, the EPS for the fiscal year 2026 was $17.28. According to CrypS.pl calculations, the price of $487.31 divided by this value gave approximately 28.2 times adjusted EPS.

This is a lower multiple than in Vertiv, but the comparison is not direct: Microsoft has a more diversified business and recurring revenues.

The biggest risk is the bill for investments. Cash paid for PP&E increased over the year from $64.6 billion to $115.9 billion.

With a consistent definition of FCF as operating cash flows minus cash paid for PP&E, the calculated free cash flow decreased by about 6.5%.

This measure does not include assets acquired through financial leasing and is not the full CAPEX presented by management.

Microsoft also indicates that demand, the degree of utilization of new infrastructure, access to energy, and components remain uncertain.

Microsoft encompasses the broadest part of the AI monetization chain from the entire list.

Investors should monitor not only Azure growth but also the relationship with OpenAI, cloud margins, free cash flows, and the utilization of new data centers.

Text S.A.: Polish AI Company or Just a Candidate? {#7}

Text, previously LiveChat Software, sells communication and customer service tools: LiveChat, ChatBot, HelpDesk, and the new Text platform.

AI agents are supposed to automate conversations, sales, and support. This is a more direct application exposure than in companies that only mention AI in their strategy.

At the same time, Text is the most conditional position on this list.

The company does not report a separate revenue segment from AI.

So far, the evidence lies in product metrics, growth in recurring revenues, and management comments, not in the income statement of the new line.

Text's Results for Fiscal Year 2025/2026 and KPI for Q1 2026/2027

In the fiscal year 2025/2026, ending March 31, 2026, Text achieved:

• 329.1 million PLN in revenue, down 7.1% year-on-year,
• 116.6 million PLN in net profit, down 29.1%,
• 153.3 million PLN EBITDA, down 23.9%,
• 83.12 million USD ARR at the end of March,
• 6.93 million USD MRR, down 2.7% year-on-year.

Details can be found in Text's annual results.

The first group KPI of the new fiscal year was better. By the end of June, MRR increased to 7.46 million USD, according to the company's announcement, up 7.6% compared to March and 4.0% year-on-year.

The group's total ARR reached 89.52 million USD, and payments increased by 10.8% year-on-year to 24.19 million USD. These are not separate KPIs for the Text platform or separately identified revenues from AI.

Caution is warranted. Text indicated that the biggest impact on the metrics was the end of the protection of old prices for LiveChat customers. A larger share of annual payments also helped.

The price migration had a retention cost. According to management, churn, or the percentage of departing customers, spiked in May and remained at a historically high level in June.

The company has not ruled out maintaining an elevated churn rate. The increase in MRR and ARR following price hikes does not therefore indicate an improvement in customer numbers or revenue sustainability.

AI Agents, Valuation, and Risk

The company reported that its AI agents achieve an average 74% case resolution rate. This is a proprietary metric: Text did not provide the full methodology, sample size, or independent audit in the announcement. The rate is a product signal, not evidence that the new platform has already changed financial results.

According to calculations by CrypS.pl, the price of 49.60 PLN on August 24 multiplied by 25.75 million shares gave a capitalization of approximately 1.28 billion PLN.

This corresponded to roughly 11.0 times net profit and 3.9 times revenue for the year 2025/2026.

Lower multiples are not automatic evidence of undervaluation.

The market is pricing in a decline in results, product transformation risk, and competition from global platforms.

LiveChat still accounted for 83.7% of revenues, and the increase in ARR was supported by price hikes. Results in PLN are also sensitive to USD/PLN.

Text must prove that AI agents increase customer numbers, usage, revenue, and profit, rather than just improving the narrative.

This is an interesting candidate from the GPW, but its thesis is less mature than that of other companies.

Not Just Nvidia. Who Really Profits from AI Development? {#8}

The most visible profits today go to infrastructure providers.

Nvidia and Broadcom show direct revenues from processors, accelerators, and networks.

Vertiv benefits from the physical expansion of data centers.

Microsoft combines infrastructure investments with rising Azure revenues and paid Copilot licenses.

At the end of the chain are applications like Text.

That’s where significant value can be created for ordinary businesses, but also where competition is fiercest, switching suppliers is easiest, and AI features can quickly become standard rather than a paid differentiator.

Therefore, the label "AI company" says little. Better questions are:

• Who pays the company for AI?
• Is the revenue recurring?
• Does AI improve margins, or does it just raise costs?
• How easily can a customer switch suppliers?
• Is the disclosed metric financial or purely product-related?

The same approach should be applied when using AI tools to support investor research.

The tool can speed up work, but it cannot replace checking the company’s report.

AI Companies on the GPW: Does Poland Have More Beneficiaries? {#9}

In 2025, 8.4% of Polish companies employing at least 10 people used artificial intelligence, compared to 20.0% across the European Union.

Eurostat data shows room for growth but does not prove that every Polish tech company is already a beneficiary of AI.

Besides Text, several companies are worth monitoring:

Data sources for the table: Medicalgorithmics, Spyrosoft, Vercom, DataWalk, Cloud Technologies, and Asseco Poland.

Medicalgorithmics has a more direct AI product than Text, but it is still a small and net loss company.

Other companies may benefit from the trend, but as of the analysis date, there is still a lack of a clear connection between AI and measurable sales.

If you are just getting to know the Polish market, a guide on how to start investing on the WSE will be helpful.

Major Investment Risks in AI Companies

1. Expenses may rise faster than revenues

New data centers generate depreciation, energy consumption, and financing costs. If demand turns out to be weaker than forecasted, some infrastructure may remain underutilized.

2. Valuations assume continued rapid growth

Good results do not guarantee stock price growth. If the market expected even better data, even a growing company may be overvalued. This particularly applies to Nvidia, Broadcom, and Vertiv.

3. A few customers account for a large part of demand

The largest laboratories and cloud companies account for a significant portion of infrastructure purchases. A shift in one project can be noticeable to the supplier.

4. Customers are developing their own chips

A large cloud may simultaneously buy Nvidia processors and develop its own accelerator. Proprietary chips do not have to displace external suppliers but can limit their pricing power.

5. Regulations, geopolitics, and supply chains

Export controls limit the sale of advanced chips. Production remains dependent on Asia, and tariffs can raise costs across the entire chain.

6. Energy and connections

A server without power does not earn. Delays in connections, permits, and the construction of power sources can postpone the launch of data centers even when processors are available.

7. Currency Risk

A Polish investor buying American stocks bears the risk of the company's exchange rate as well as USD/PLN. A strengthening zloty can reduce returns in PLN despite an increase in stock prices in dollars.

8. Product Metrics Are Not Revenue

The number of agents, inquiries, or resolved cases can be useful, but it does not replace revenue, margin, and cash. This is particularly important for smaller application companies.

Is it too late to invest in AI by 2026? {#11}

It cannot be answered solely based on the price chart. An increase in price does not prove overvaluation, and a decrease does not indicate an opportunity. It is better to prepare scenarios.

• Bullish Scenario assumes that the expenditures of the largest clouds remain high, demand exceeds supply, and applications increase the use of infrastructure. Then revenues can grow at several levels of the chain simultaneously.
• Base Scenario means further development of AI, but a slower pace of orders and normalization of margins. In such an environment, valuation, diversification, and cash generation capability become more important.
• Bearish Scenario assumes a delay in return on investment, cuts in expenditures, excess capacity, or price pressure. Companies whose stock price assumes years of growth without setbacks may be the most vulnerable.

There is no need to choose a single company. An alternative is AI ETFs.

The fund limits the risk of a single company but may be concentrated in a few largest positions and include companies loosely related to AI.

AI Companies 2026: Watchlist Table