Beyond the Hype: The 4 Biggest Takeaways from CES 2025
#1: Physical AI is Primed for Development, but ROI Remains Elusive
NVIDIA’s CEO Jenson Huang commenced CES 2025 with the opening keynote speech that revealed several pioneering AI announcements from the technology giant. One of the key announcements from NVIDIA was the unveiling of its physical AI platform, NVIDIA Cosmos. Physical AI is defined by NVIDIA as AI that extends current generative AI models by having an awareness and understanding of the spatial relationships and physical behaviour of a 3D or physical world. Physical AI will enable the use of autonomous machines, including robots and self-driving cars.
Source: NVIDIA
NVIDIA Cosmos will come with state-of-the-art generative World Foundational Models (WFMs), advanced tokenisers, guardrails and an accelerated video processing pipeline that will aim to accelerate the development of physical AI systems. However, large amounts of data are required to train physical AI systems to successfully navigate a physical world, and NVIDIA’s WFM aims to provide an effortless way to generate substantial amounts of physics-based synthetic data that will facilitate the fast training and evaluation of these physical AI systems. The platform will streamline the process of training; making it more cost and time-effective for end users.
While physical AI is expected to optimise several applications, including self-driving cars, traffic-control systems and surgical rooms, NVIDIA’s Vice President of Omniverse and simulation technology, Rev Lebaredian, identified the industrial sector and predicted that it will bring AI to 10 million factories and 200,000 warehouses.
In addition, NVIDIA announced that they would be furthering its support for agentic AI, with new AI blueprints that simplify the development of AI agents. Agentic AI consists of multiple AI agents that work together to independently set and meet goals based on user input, making them capable of solving complex, multi-layer problems. Whereas generative AI generates results based on patterns it learns in training, agentic AI can generate results by understanding both the context and goal of a task. In collaboration with leading agentic AI providers, including CrewAI, Daily, LangChain, LlamaIndex, and Weights & Biases, NVIDIA’s new AI blueprints integrate NVIDIA’S AI enterprise software platform, NVIDIA NIM microservices and NVIDIA NeMo.
Source: NVIDIA
These announcements usher in a new era of AI models with greater contextual understanding that are set to significantly disrupt how enterprises operate, allowing for the automation of key tasks with higher accuracy. However, given the nascency of these two technologies, there are several obstacles that the market must overcome before the widespread adoption of both physical and agentic AI.
- Despite NVIDIVA’s efforts to streamline training for physical AI systems, deployment cost will remain substantial and integration into existing systems will remain complex.
Vendors looking to deploy physical AI or agentic AI must focus on providing optimised models that are less resource-intensive and sustainable, allowing for lower maintenance costs. This will make these two technologies accessible to a wider audience and allow for the widespread adoption of both agentic and physical AI in the long run.
- Secondly, to build market confidence, vendors deploying physical AI or agentic AI must cultivate portfolios of case studies by partnering with large industry leaders.
This will not only instill confidence in potential users from real-world examples but will also allow them to acquire industry-specific expertise that will allow them to offer unique and value-added services which will allow them to further penetrate the market.
Juniper Research anticipates that, over the next year, large industry leaders will experiment by deploying the technology in sandbox conditions so as to not disrupt well-refined revenue-generating processes. The key to widescale adoption now is proof of return on investment (ROI) – without this, the technology will never progress.
#2: The Increasing Need for Tech Partnerships in Automotive
Amazon announced several key autonomous vehicle partnerships at CES 2025; solidifying the technology giant’s presence in the autonomous vehicle market. One of the key partnerships it announced was a ten-year cloud infrastructure agreement between its subsidiary Amazon Web Services (AWS) and HERE technologies, a Dutch multinational group that specialises in mapping technologies and location data.
The partnership will aim to provide a streamlined solution for the development of location-aware technologies, which will further enhance advanced driver-assistance systems (ADAS) and automated driving (AD) for the automotive, logistics and mobility companies worldwide.
Source: HERE
The partnership will achieve this by combining AWS’ cloud expertise and HERE’s location intelligence to support the automotive industry in developing and deploying AI-powered, live streaming location services in real-world applications.
It will also address key challenges in deploying ADAS and AD solutions in the automotive industry by providing services that accelerate both the testing and real-world deployment of ADAS and AD solutions. For example, automative developers can now easily locate and deploy the HERE HD map into simulations for testing, using both the HERE HD Live Map in conjunction with AWS’ AI tools such as Bedrock.
In addition to this partnership, AWS has agreed a similar partnership with French global automotive supplier, Valeo. The partnership will entail Valeo offering new solutions, powered by AWS, which will aid the development process of software-defined vehicles (SDVs), allowing for faster, cost-efficient deployments of SDV technologies.
Furthermore, Amazon has also entered into an automobile technology partnership with Qualcomm to develop advanced in-vehicle experiences. The partnership will see Qualcomm’s automotive expertise combined with Amazon’s AI capabilities to advance personalised and cutting-edge driving experiences; allowing automobile manufacturers a faster, flexible and cost-effective method to develop advanced infotainment and AI features. It will also aim to provide a virtual environment for automobile manufacturers to develop and test infotainment and driver-assistant technologies.
Amazon will also start rolling out large language model (LLM)-powered capabilities with its partner BMW. This follows the partner’s successful roll-out of the new BMW Intelligent Personal Assistant (IPA), powered by Amazon’s Alexa Custom Assistant (ACA) technology, in BMW’s X3 model in July 2024.
Source: Amazon
The partnerships of both Amazon and AWS with automotive companies and automotive technology companies highlight the shift of the automobile market to an intelligent, software-centric and cloud-connected future. The navigational advancements that this shift will bring will not only enhance the automotive industry, but the logistics industry as well; providing more efficient supply chain management. Amazon’s key partnerships will accelerate the development of SDVs, however we foresee several key hurdles that Amazon and its partners, as well as other competitors in this industry, will have to overcome to successfully deploy these technologies.
The biggest challenge will be regulatory hurdles and gaining market confidence. Similarly to the physical and agentic AI market, these hurdles can be overcome by forming partnerships and collaborations with regulatory bodies, such as WP.29. These collaborations must involve streamlined certification processes that balance thorough testing with efficient deployment roll-outs, and promote collaboration.
Cybersecurity will also be a key concern, and technology and automobile companies must also work with regulatory bodies to create standards and certification programmes to enhance the deployment of these technologies, as well as to increase market trust. Lastly, the shift to SDVs will also provide a unique workforce obstacle where the existing workforce will need to be reskilled to support this shift. Automobile and technology vendors in this industry must invest in educational programmes and workshops to support their workforce through this transition.
#3: Questions Raised over Quantum Development Timelines & ROI
When questioned by analysts on the outlook for quantum computers, NVIDIA CEO Jensen Huang implied that quantum technology was at least 20 years away; claiming that a return on investment in 15 years would be considered early, and in 30 years, late. This resulted in a plunge in stock of key quantum technology leaders, including IonQ, Rigetti, Quantum Computing Inc and D-Wave Quantum, with shares crashing by almost 40%.
While quantum technology is an exciting prospect, particularly in advanced optimisation solutions that will transform sectors with complex models, such as finance and logistics, there are key challenges which are inhibiting growth in the quantum market. The nature of qubits, which are bits of quantum information, require cryogenic systems that are expensive to operate. In addition, qubits suffer from quantum decoherence; a phenomenon where they lose their quantum state and convert to a classical state due to interactions with their environment.
This can lead to information loss and errors. Quantum error correction methods, which use algorithms to protect quantum information by redundantly encoding quantum information across several qubits, can be used to mitigate these errors, but these methods can be computationally expensive to implement successfully. These two factors therefore add to the high cost of maintaining a quantum computer, let alone purchasing one, making quantum technology difficult to access for substantial portions of the market.
While these hybrid quantum classical computers cannot perform on the same level as quantum computers, they currently provide an assured method of achieving optimised quantum computing in real-world scenarios, without the challenges quantum and supercomputers currently face.
A good example of is this the solution that Q-CTRL provided for the Australian Army’s convoy in February 2024. Its task was to find the most optimised route for travelling, saving costs and time. Previously, quantum computing solutions had been an issue, due to error detection and correction, as well as noise. However, quantum company Q-CTRL deployed a hybrid solution, which was able to provide a highly optimised route for the army.
The popularity of offering quantum-as-a-service (QaaS) is also growing; allowing companies to access quantum computing without the high implementation and maintenance costs. Although prices remain high compared to other ‘as-a-service’ business models, it has greatly increased accessibility of quantum computers.
These two solutions, combined with large government investments particularly from countries such as the US, China, the UK, Japan, South Korea, Germany and France, will drive a large amount of innovation in the quantum technology market. Juniper Research anticipates that global cumulative investments will reach $45 billion by the end of 2025; a 139% increase from 2023. This will be driven by a quantum technology race between the key leading countries mentioned above, with governments focused on ensuring that they remain up to date on quantum innovations, particularly those relating to national security.
Although quantum technology is physically fragile, and expensive to implement and maintain, its development is similar to AI, which has seen increasing investment over the last decade and is now considered a well-established and well-defined technology. Depending on the model, AI can be computationally expensive and time consuming to train to start obtaining highly precise results.
#4: IoT to Focus on Service Simplicity for SMEs to Achieve Growth
The computer intelligence company Qualcomm announced the upgrade of its cloud-based Qualcomm Aware platform. The upgrade will include new key features that allow companies to include observability, monitoring and location capabilities to intelligently connected IoT devices in industries such as energy and logistics.
Source: Qualcomm
The additional features of the horizontal IoT enablement platform aim to tackle some of the key challenges in building; customising and deploying intelligent connected solutions. The features include observability and insights, geolocation, fine indoor positioning, firmware updates and device management services. This will allow device manufacturers and original equipment manufacturers (OEMs) to gain enhanced intelligence into connected devices which will allow them to troubleshoot devices as well as take preventative measures to fix issues that are observed. This will also enable OEMs and device manufacturers to gain insights into delivery efficiency and allow users to ensure the longevity of devices by easily keeping them up to date on the latest firmware and software updates.
Given the global increasing digital transformation and the increasing number of connected devices, enterprises are rapidly adopting IoT solutions, which themselves are rapidly evolving. However, these solutions are also growing in complexity, and security becomes a significant concern as many IoT devices may not be properly secured.
The new features that Qualcomm has added will allow enterprises and industries to gain industry-specific insights and increase innovation. In addition, these features will allow device manufacturers and OEMs to refine their products and software, allowing for more enhanced visibility into IoT infrastructures and simplifying previous complex IoT ecosystems. Juniper Research anticipates that the total number of connected IoT devices will reach 21 billion by the end of 2025. These solutions will be especially valuable to small to medium-sized enterprises (SMEs) who may not have the required level of expertise to implement and manage the growing number of connected devices.
Total Number of IoT Devices (bn), 2024-2029
Source: Juniper Research
As such, Juniper Research anticipates that platforms which provide greater insights into IoT devices will be key to ensuring both secure and enhanced IoT solutions, as well as streamlining the manufacturing and deployment processes of IoT devices.
As a Research Analyst within Juniper Research's IoT & Emerging Technology team, Michelle provides insight and analysis on the latest developments in nascent, fast-growing technology markets. Her latest reports include IoT Cybersecurity, Advanced Air Mobility, and Quantum Technology.
Latest research, whitepapers & press releases
-
ReportSeptember 2025Fintech & Payments
Instant Payments Market: 2025-2030
Juniper Research’s Instant Payments research suite provides a wide-ranging and strategic analysis of this market; enabling stakeholders - from banks, infrastructure providers, regulators, and businesses - to understand future growth, key trends, and the competitive environment.
VIEW -
ReportSeptember 2025Fintech & Payments
Anti-money Laundering Systems Market: 2025-2030
Our AML Systems research suite provides a detailed and insightful analysis of this evolving market; enabling stakeholders from financial institutions, law enforcement agencies, regulatory bodies and technology vendors to understand future growth, key trends, and the competitive environment.
VIEW -
ReportSeptember 2025Fintech & Payments
A2A Payments Market: 2025-2030
Our A2A Payments research suite provides detailed analysis of this rapidly changing market; enabling A2A payments service providers to gain an understanding of key payment trends and challenges, potential growth opportunities, and the competitive environment.
VIEW -
ReportSeptember 2025Telecoms & Connectivity
Mobile Messaging Fraud Prevention Market: 2025-2030
Our Mobile Messaging Fraud Prevention research suite provides a detailed and insightful analysis of a market set for significant disruption over the next five years. It enables stakeholders from mobile operators, enterprises, and mobile messaging fraud prevention vendors to understand how the market for mobile messaging fraud will evolve, as well as the impact of AI, RCS, and the evolving competitive environment.
VIEW -
ReportSeptember 2025Sustainability & Smart Cities
Smart Grid Market: 2025-2030
Our cutting-edge Smart Grid research suite provides a comprehensive view of a market at the forefront of the global energy transition. It examines the major disruptions transforming the sector, from the integration of distributed energy resources and the rise of virtual power plants to the growing role of AI-driven intelligence and the mounting need for robust cyber security and compliance.
VIEW -
ReportSeptember 2025Fintech & Payments
eCommerce Payments Market: 2025-2030
Juniper Research’s eCommerce Payments research suite provides a comprehensive and insightful analysis of this market; enabling stakeholders, from eCommerce payment platform providers to merchants and payment service providers, to understand future growth, key trends, and the competitive environment.
VIEW
-
WhitepaperSeptember 2025Fintech & Payments
Core Banking Transformation - A Strategic Conversation with SAP Fioneer
Core banking transformation is no longer optional, as regulatory change, rising compliance costs, and shifting customer expectations make legacy systems unsustainable. Anna Koritz, Global Head of Transaction Banking at SAP Fioneer, shares how banks can overcome cultural and technical hurdles and why SAP Fioneer’s modular, cloud-ready approach enables confident modernisation.
VIEW -
WhitepaperSeptember 2025Fintech & Payments
From Detection to Prevention: The Next Era of Anti-money Laundering
Our complimentary whitepaper, From Detection to Prevention: The Next Era of Anti-money Laundering, examines the state of the AML systems market; considering the impact that a changing regulatory environment and a growing number of use cases is having on the market. Additionally, it includes a forecast summary of the total value of the AML systems market in 2030.
VIEW -
WhitepaperSeptember 2025Fintech & Payments
Ascending-to-Ailing: The Deceleration of A2A Adoption
Our complimentary whitepaper, Ascending-to-Ailing: The Deceleration of A2A Adoption, examines the state of the A2A payments market; considering the impact of this payment method and how it is shaping the modern payments landscape through lower fees and enriched user experience.
VIEW -
WhitepaperSeptember 2025Telecoms & Connectivity
RCS Fraud: Emerging Threats in Next-gen Messaging
Our complimentary whitepaper, RCS Fraud: Emerging Threats in Next-gen Messaging, examines the future of the messaging fraud prevention market, with a particular focus on the latest trends within RCS Business Messaging (RBM). Additionally, it includes a forecast summary of the total cost of fraud over RBM to subscribers in 2030.
VIEW -
WhitepaperSeptember 2025
Decentralising the Smart Grid: Opportunities & Challenges
Our complimentary whitepaper, Decentralising the Smart Grid: Opportunities & Challenges, explores how distributed energy resources, renewable integration, and virtual power plants are reshaping grid management.
VIEW -
WhitepaperSeptember 2025Fintech & Payments
Going Glocal ~ Why Local Payment Methods Are Driving eCommerce
Our complimentary whitepaper, Going Glocal ~ Why Local Payment Methods Are Driving eCommerce, assesses how local payment methods are driving the increasing accessibility to eCommerce, and challenges to eCommerce growth.
VIEW
-
Fintech & Payments
AML Systems Market to Surpass $75 Billion by 2030 Globally, With LexisNexis Risk Solutions, Oracle, and Experian Leading the Defence
September 2025 -
Fintech & Payments
B2B Payments to Hit $224 Trillion by 2030 Globally, Driven by Emerging Market Expansion
September 2025 -
Fintech & Payments
A2A Transaction Value to Reach $195 Trillion in 2030 Globally, Driven by Advanced Value-added Services
September 2025 -
Telecoms & Connectivity
ReveNet: Operators Must Act to Restore Trust & Transparency to $55bn A2P SMS Ecosystem
September 2025 -
Telecoms & Connectivity
RCS Business Messaging Fraud to Cost Mobile Subscribers $4.3 Billion Globally Over the Next 5 Years
September 2025 -
Sustainability & Smart Cities
Smart Grids to Support 43% of Global Electricity Supply by 2030, Driven by Virtual Power Plants
September 2025