• Stronger inflation, coupled with solid payroll growth, will likely cause the Federal Reserve to raise interest rates by the end of this year. However, the rates are unlikely to stay heightened for long. We expect the Fed to announce a rate cut before the end of 2027. We expect oil prices to move lower next year, which will cause inflation to move lower on a sequential basis.
• Consumer spending is also likely to shift lower amid higher interest rates and inflation. Indeed, personal savings fell to an extreme low of 2.6% of after-tax income in April. Wage growth has also continued to moderate even as inflation has picked up, further eroding purchasing power. Wage growth is not expected to pick up either, given demand for labor remains muted. A decrease in immigration is expected to weigh on slow total employment growth, with working age population growth nearing 0%, slowing to a crawl.
• Real GDP is still expected to grow by a healthy 2.0% in 2026, before slipping to 1.8% in 2027, when consumer spending is on a more sustainable footing and AI-related investment growth moderates a bit. However, we have revised our longer-term real GDP forecast higher as we now anticipate stronger AI-led productivity gains to begin showing up in macroeconomic data over the next few years. As a result, real GDP in 2030 is expected to be 2.1%, up from the 1.7% we forecasted in March.
• We assume that Brent crude averages US$106 per barrel in 2026, slipping to a still-lofty US$103 in 2027. We also assume that AI investment gets overdone, leading to a sharp pullback in business spending in 2027, as companies reassess potential demand for related products
• As a result, we expect real consumer spending to remain unchanged in 2027 and fall by 0.4% in 2028.
• The weakening of domestic demand raises the unemployment rate to 6.5% in 2028. It also provides a stronger disinflationary impulse, allowing growth in core personal consumption expenditure (PCE) prices to dip below the US Fed’s 2% target by the end of 2027, before returning to 2% in 2030.

4 more annotations...

• According to the Bureau of Labor Statistics, the Consumer Price Index showed inflation moving from 2.4% in January and February 2026, to 3.3% in March, and a whopping 3.8% in April.  People have to eat and drive, and rising costs reduce households’ ability to pay off their credit obligations. Credit cards haven’t become more expensive than they were when consumers took them out, but consumers’ budgets aren’t keeping up with the financial demands.
• This year will likely end with slightly higher credit losses as economic events continue to play out.  Watch out for 2027, though. Loan loss reserves for 2026 are already locked in and shown to be protected by DFAST. It is time to think about your 2027 MBOs—and in that, anticipate higher credit losses in late 2026, rolling into 2027.

• In the United States, banks are reducing cross-border lending to higher exposure countries while maintaining activity via foreign subsidiaries, limiting potential losses in extreme shocks, yet preserving contagion channels by shifting part of the risk onto domestic balance sheets.
• In February 2026, the United States Treasury published two key references for AI¹⁴ governance in financial services: the Financial Services AI Risk Management Framework (FS-AI RMF) and the AI Lexicon. Both translate into operational obligations principles already demanded by supervisors: explainability, governance, traceability, accountability, and lifecycle control.
• Institutions must evidence how all AI-driven decisions are generated, supervised, validated, and documented. “Pure outputs” from models such as GPT¹⁵ or Claude¹⁶, without human review, full traceability, and technical reproducibility, are no longer acceptable from an internal control and compliance standpoint. This requires embedding human-in-the-loop (HITL¹⁷) oversight, strengthening end-to-end auditability, ensuring data provenance and integrity, and assigning formal owners for each use case.
• In 2026, the banking sector operates in a more uncertain, asymmetric, and demanding environment than in previous years. Persistent geopolitical tensions, supply-side shocks, and institutional renewal at leading monetary authorities are accelerating existing dynamics and increasing the likelihood of extreme events.
• The expected rise in interest rates provides banks with a unique and temporary window of opportunity: profitability is higher, and capital and liquidity positions remain strong. This makes the current moment a “now or never” juncture to invest decisively in digitalization and AI, particularly considering growing competition from FinTech²⁹ that are starting to become profitable.

3 more annotations...

• Global growth is likely to slow modestly to 3.2% for the full year, hurt by the energy shock but avoiding recession, and then recover to 3.4% in 2027 as oil and gas prices ease.
• AI-related spending is the dominant force in the current investment cycle—and critical to the resilient U.S. growth outlook. Business spending in the U.S. should rise 7% in the fourth quarter from a year earlier and 8% in 2027 overall.
• In the U.S., higher oil prices are being offset by a lessening in the impact of tariff-related inflation in the second half. By the end of this year, U.S. inflation should begin to get back to the modest downward trend that was seen at the start of 2026.

• In Morgan Stanley’s baseline forecast, the Fed will cut rates twice in the first half of 2027, by 25 basis points in the meetings of March and June to a final range of 3% to 3.25%, as inflation begins to recede.

   

   

2 more annotations...

• Faiyaz Elahi Mullick, an ECE doctoral candidate, is experimenting with edge computing systems, which conserve massive amounts of energy by processing data on small devices like smartphones instead of sending data to large cloud servers, which are typically housed in data centers.

  His research centers around computing systems that mimic the mammalian brain’s ability to remember.

• As it stands, many AI systems struggle with a problem known as catastrophic forgetting. Sometimes, when you train an AI system to recognize dogs, for example, and then teach it to recognize a plant, it will overwrite its memory and forget what it learned about dogs. Humans and other mammals, on the other hand, can learn new things throughout their lives without automatically erasing old memories.

  Researchers are exploring whether a biological mechanism known as adaptive synaptogenesis, which helps the brain form new connections while conserving energy, could help AI systems do the same.

• The global AI market is expanding rapidly, projected to grow from USD 189 billion in 2023 to nearly USD 5 trillion by 2033. This would represent roughly a 25-fold increase in global AI market size over a decade. Global AI expenditure is projected to exceed USD 2.5 trillion in 2026.
• Nearly half of the world’s data centers are in the United States. Only 16% of countries host AI-specialized cloud compute, and 90% of that capacity is concentrated in just 2 countries (the United States and China).
• The training of frontier models demands immense energy . GPT -4 likely consumed 50 to 70 GWh of electricity over 100 days, roughly 40–55 times more than GPT-3 (1.287 GWh over a 34-day period). GPT-4’s training energy is equivalent to the annual residential electricity consumption of over 460,000 people in Sub-Saharan Africa.
• GPT-4’s training carbon footprint of 25,000 tonnes of CO₂e would require the sequestration capacity of 420,000 tree seedlings grown for 10 years, or about equal to the number of trees in 105 Hyde Parks in London.
• The water footprint associated with training GPT-4 was about 600 million liters, enough to meet the minimum annual domestic water needs of 81,000 people in Sub-Saharan Africa, or to fill 237 Olympic-sized pools.
• The water footprint of GPT-5 training is estimated at 1 billion liters, enough to meet the annual domestic water needs of more than 135,000 people in SubSaharan Africa.
• While training is highly resource-intensive, the continuous inference phase used to generate responses for billions of interactions is estimated to account for 80% to 90% of total AI energy use.
• Generating a typical AI image requires 2.9 Wh, making it 60 times more demanding than a short text answer and 1,450 times that of text classification.
• In 2025, global data centers were estimated to consume 448 TWh of electricity. If data centers’ electricity use were considered a country, it would have ranked 11th globally by electricity consumption.
• The energy consumed by data centers in 2025 was enough to supply the annual residential electricity needs of the entire population of Sub-Saharan Africa, 1.3 billion people, for 2.6 years.
• By 2030, AI infrastructure could generate up to 2.5 million metric tons of e-waste each year, roughly equivalent to discarding 250 Eiffel Towers annually.

9 more annotations...

• The challenges to the industry come at a time when the global protection gap (the difference between actual and insured losses), which reached US$1.4tn in 2020, is widening at an accelerating pace, according to PwC research. PwC’s analysis estimates this gap could reach US$1.86tn by 2025, with the Asia-Pacific region accounting for almost half of all uninsured risk.
• For insurers, this presents a significant opportunity to expand their market and bridge the protection gap through innovative products and services. At the same time, it also highlights the need for insurers to adapt and evolve in order to effectively reach underinsured customers and provide them with relevant coverage options.
• A significant shift is anticipated from a 'promise to pay' model to a 'promise to help' model, particularly in health and car insurance sectors. This entails evolving from merely compensating for losses to encouraging proactive preventative measures
• nsurers will need to collaborate with other industries, such as healthcare and technology, to offer customers personalized solutions that not only protect against potential risks but also promote healthier lifestyles
• The insurance sector is expected to increasingly leverage digital technologies to enhance customer experience. This includes engaging in digital ecosystems, building unique customer intelligence, and designing products and services to meet the demands of a digital world.
• This can include automating routine tasks, utilizing chatbots for customer inquiries, and developing personalized pricing models based on real-time data.
• Accelerating technology trends will transform the nature of insurance, with auto insurance risk shifting from drivers to the AI and software behind self-driving cars, and real-time data sets giving insurers unprecedented visibility into risks​.
• Insurers will need to adapt to the digital and AI-driven landscape by incorporating new technologies into their operations. This includes phasing out outdated systems and implementing emerging tech solutions such as no-code platforms

6 more annotations...

• significant majority (86% for security and 70% for governance) of respondents prioritize security and governance above all else. Our view is this inclination tends to favor more integrated platforms like Snowflake, which require customers to put their data into Snowflake to take advantage of the most comprehensive governance solutions.
• Conversely, a substantial cohort is focused on avoiding vendor lock-in at all costs, aligning more with Databricks’ open-source ethos
• Snowflake has cemented its place as the leader in cloud DBMS, while Databricks has dominated the data pipeline space with Spark and other tooling.
• • While still early days, the interest in interest in adopting open table formats, particularly Iceberg, is on the rise, with 70% of respondents indicating a shift towards this format.
      
• The governance layer is becoming the new strategic control point, moving beyond traditional DBMS. Key players are attempting to make this the new ‘moat’ in our view. This includes Databricks’ Unity Catalog, Snowflake’s Polaris, which must coexist with a variety of solutions from Google, Microsoft, AWS, Informatica, Collibra, Alation and others. The governance landscape remains highly fragmented, with a plethora of solutions and a complex ecosystem of partnerships and standards

3 more annotations...

• rail is fundamentally enhancing the technical core of the Dynatrace Software Intelligence Platform.
• rail was purpose-built with one fundamental goal: making all data accessible—and thus valuable—to provide precise answers in real-time, boost insights gathered by Davis AI and drive automation. This goal isn’t limited to observability efforts. Grail needs to support security data as well as business analytics data and use cases.

• Grail needs to achieve three main goals with minimal impact to cost:

   
   
  • Cope with and manage an enormous amount of data—both on ingest and analytics
   
  • Work with different and independent data types
   
  • Put data in context and enrich it with topology metadata
• we had to decouple storage from analytics and compute. This decoupling ensures the openness of data and storage formats, while also preserving data in context. Further, it builds a rich analytics layer powered by Dynatrace causational artificial intelligence, Davis® AI, and creates a query engine that offers insights at unmatched speed.
• customers can choose between Dynatrace® OneAgent or open source observability frameworks, such as OpenTelemetry, Prometheus, Micrometer, and others.
• Dynatrace OneAgent is the only agent that automatically collects data from every tier of your application stack without requiring configuration. Just a few minutes after installation, you get all the performance metrics and log data you need to monitor IT infrastructure of any complexity—
• Grail enables highly efficient processing pipelines to enrich ingested data, extract metrics, or add a privacy filter.

• Once ingested, Grail ensures that data is retained for up to three years – fully flexible, based on your needs and individual settings – while still being accessible by the analytics layer. Data is available in real time without requiring indexing by our powerful Dynatrace Query Language.

   
• Schema-on-write worked well for some time, but it has reached its limits. The concept has major disadvantages that are significant when considering today’s ever-growing data complexity and quantity. Data is either not available at all, or you need to invest a lot of effort and wait for hours, or even days, until it is re-indexed.
• Grail’s schema-on-read allows IT teams to store data in its native format. That means data preparation occurs when data is queried, not ingested. This provides greater flexibility, as teams can perform any query at any time—on data in its native structure—and are not limited to the fields defined in the schema when the ingest pipeline was set up.
• Grail is also indexless. Whereas indexes promise high-query performance with indexed data, they’re limited in size due to high costs. As a result, data is generally stripped—or dropped—at ingest, to minimize the size of the index. Additionally, data is moved from index to cold storage once the data reaches a certain age.
• Thanks to its massively parallel processing (MPP) engine, you can perform any query and retrieve results instantly. Use the power of the Dynatrace Query Language (DQL) to search, parse, filter, sort and aggregate all data stored and available within Grail

10 more annotations...

• From day one, Grail disrupted the log management and analytics market by unifying observability, security, and business data and providing instant answers thanks to its massively parallel processing (MPP) capabilities.   Further extending our platform's analytics capabilities, we're increasing Grail's capabilities by adding new data types and unlocking support for graph analytics
• odern cloud-native architectures can result in scalability and fragmentation challenges. Ensuring observability across these environments requires access to data at a massive scale. The proliferation of metrics can quickly result in a high cardinality challenge, with each service, host, or Kubernetes pod adding its own unique values to the data set.
• Metrics on Grail is architected to manage billions of metrics to cope with cardinalities and unique value combinations of 1 trillion potential permutations for timeframes beyond a year. This is only possible because of our no-index approach and massive parallel processing capabilities, which enable Dynatrace to offer extra-long data retention (15+ months) at full granularity that is cost-efficient and fast.

• With Grail, we address these customer challenges by offering the most powerful and future-proof trace analytics solution on the market, which:

   
   
  • Handles data volumes of hundreds of terabytes a day
   
  • Retains large data volumes for up to 15 months in a highly cost-efficient way
   
  • Ensures that data retains its context by assembling trace spans into PurePath® distributed traces (including additional code and thread profiling data)
   
  • Returns instant query results in real-time using indexless queries
• With Dynatrace and Smartscape for DQL, metrics are a completely different game. Whether it’s metrics, logs, events, traces, or any other data type, Dynatrace not only retains the data context but also enables you to analyze data in its semantic context without boundaries.
• These capabilities are powered by Smartscape for DQL, a directional graph representing the real-time topology and dependencies of a data architecture. Smartscape unifies the different data types ingested into Dynatrace and retains the full context of this data to enable holistic and precise data analytics.
• With the Dynatrace Query Language (DQL), teams can perform these analyses by asking questions that weren’t possible in the past. There are now boundless possibilities, such as identifying users affected by a service outage in a red-alert scenario or doing forensic research on a recent data breach.
• When we started working on Strato—the new Dynatrace design language that powers our new user experience—we developed a few core principles to address the design challenges stated above
• the interface is user-centric rather than data-centric.
• The interface is simple
• The interface offers infinite possibilities
• Dynatrace® Dashboards transforms complex data into easy-to-understand visualizations. Dashboards is your go-to app for quick and clear data overviews, whether you need a status-quo view that can be observed over time or you need to share aggregated views with management or business teams
• Dynatrace Notebooks is your on-demand window into data exploration. It addresses the challenge of finding the right data, cleaning, filtering and transforming the data and finally connecting with other data to understand underlying dependencies
• Start creating data-driven documents and perform custom analytics. Depending on your use case, Notebooks can persist a status quo and create a snapshot whenever necessary or be “self-updating” using current data to always reflect the actual status. You can easily interact with any query result by “slicing and dicing” the data stored in Grail: advanced filters, refinements, aggregations, and sort orders, are just a click away
• , Dynatrace users can now perform any custom query, leveraging Dynatrace Grail AI-fueled graph analytics power.
• • quickly identify impacted users by mapping observability and security findings to your business context and improve customer satisfaction by querying for e-commerce customers who are unable to finalize their check-outs due to a service outage.
   
  • Automation: understand the potential impact of remediation actions on dependent components.
   
  • Security: protect customers and brands by conducting application security forensics to identify, mitigate, and prevent data breaches.
   
  • Business process health: show and analyze the health status of complex processes even if dependencies are non-transactional.
   
  • Optimize: enable more efficient multicloud operations by predicting cloud performance and utilization over time to optimize resource allocation based on user needs.

14 more annotations...