Introduction: As AI Computing Power Grows, the System's Demand for "Time" Actually Increases
Over the past few years, discussions on AI hardware have almost exclusively revolved around GPUs, HBM, Advanced Packaging, high-speed interconnects, and power systems. As the computing power of a single machine steadily improves and the scale of server clusters continues to expand, time—a seemingly fundamental issue—has become increasingly critical: How does the system maintain a continuous time reference when a server loses power, the main controller restarts, the network is interrupted, or the device enters deep sleep? Furthermore, how can fault logs, temperature alerts, power anomalies, and software upgrade records accurately reconstruct the chronological sequence of events?
An RTC (Real-Time Clock) chip is a low-power chip that can continue to keep time relying on a backup power source when the device is powered off or in sleep mode, providing the system with date, time, timed wake-up, and event timestamps. The most critical component of an RTC is the crystal oscillator, which typically provides a stable 32.768kHz clock reference; some RTCs require an external crystal oscillator, while others integrate the crystal oscillator or a temperature-compensated crystal oscillator (TCXO) directly within the package.
In AI servers, high-speed data transmission relies on high-frequency oscillators, clock generators, and synchronization chips. The RTC is not responsible for clock synchronization among GPUs, optical modules, or high-speed interfaces. Instead, it undertakes another, more fundamental task—continuing to keep time after the main system stops running, preserving continuous time information for the BMC (Baseboard Management Controller), motherboard firmware, system logs, and security audits. Data provided by users indicates that the evolution of AI servers and communication equipment is simultaneously raising the requirements for timing devices in terms of accuracy and supply scale. High-precision oscillators have been applied to optical modules in AI servers, and relevant manufacturers are also advancing the development of higher-precision products and capacity building.
Similar demands are also spreading to edge AI devices, robots, and smart vehicles. Edge terminals need to use the RTC to periodically wake up the main controller to reduce long-term standby power consumption; robots and vehicles generate massive amounts of sensing, control, and fault data, requiring continuous timestamps to support post-event tracing; security systems must rely on trusted time to verify certificate validity, identify log rollbacks, and prevent replay attacks.
AI has not turned the RTC into a high-computing-power chip, but it has made the system more dependent on accurate, continuous, and trustworthy time. Computing power determines how much data a device can process, while the RTC helps the system answer another question: After power loss, sleep, and failures occur, can the device still accurately know what happened previously?
What Kind of RTC Chips Are Needed in the AI Era?
The AI system's requirements for time are not limited to high frequency and high precision. GPUs, optical modules, and high-speed interconnects focus more on low jitter, multi-node synchronization, and high-speed data transmission; the RTC undertakes another fundamental task: continuing to maintain continuous, stable, and traceable local time when the main system loses power, enters sleep mode, resets, or loses network connectivity.
In AI servers, the RTC typically serves the BMC, motherboard firmware, and system management modules. During normal system operation, network time can be obtained via NTP (Network Time Protocol) and PTP (Precision Time Protocol); when network disconnection, restarts, or power anomalies occur, the local RTC is responsible for maintaining time continuity. Events such as temperature alerts, GPU drop-offs, storage failures, and software upgrades all require reliable timestamps to reconstruct their chronological sequence. The larger the server scale and the more complex the hardware and software architecture, the higher the cost of fault localization and security audits caused by disordered time records.
Edge AI devices place greater emphasis on the low-power capability of the RTC. Devices such as smart cameras, AI door locks, and environmental sensors are typically woken up periodically by the RTC to the main controller to complete data collection, inference, and communication before returning to deep sleep. At this point, the timing current directly affects battery life. In projects where the main controller needs to be completely powered off, the device operates offline for long periods, or high precision is required, standalone RTCs still hold significant value.
Robotics, automotive, and industrial AI, on the other hand, focus more on the independence of the time source. BMS (Battery Management Systems), T-Box, EDR (Event Data Recorders), and industrial recording devices continuously generate operational status, control commands, and fault alerts. Once the time relies entirely on the main controller, main controller resets, software crashes, or power domain anomalies could corrupt critical logs. Therefore, high-reliability systems often retain a standalone RTC to provide a relatively independent time basis for fault tracing and liability determination.
Security requirements are also changing the product positioning of RTCs. Digital certificates, remote upgrades, software licenses, and audit logs all depend on the correct time. If the system time is maliciously rolled back, it could lead to invalid certificate judgments, replay of historical commands, or disruption of log sequences. Consequently, high-end RTCs are beginning to incorporate tamper detection, secure timestamps, time modification logs, and monotonic counters, and are collaborating with TPMs (Trusted Platform Modules), TEEs (Trusted Execution Environments), or secure MCUs (Microcontroller Units), gradually becoming part of trusted systems.
The future technological evolution of RTCs mainly focuses on four directions. First, nano-amp level power consumption and smarter power management, switching between high-precision and low-power modes based on system status. Second, further integration of crystals, RTCs, and temperature compensation; built-in crystals, TCXOs, and DTCXOs can reduce uncertainties caused by device matching, PCB (Printed Circuit Board) parasitic parameters, and production calibration. Third, transitioning from ordinary timekeeping to trusted time, enhancing the trustworthiness of time data through protected registers, non-erasable logs, and hardware security mechanisms. Fourth, collaborating with system software to build a hierarchical time system jointly composed of network time, on-chip RTCs, standalone RTCs, and high-speed synchronous clocks.
It is difficult for RTCs to replicate the growth curves of GPUs and HBM, and basic functions will continue to be absorbed by MCUs and SoCs (System on Chip). However, as AI enters automotive, industrial, energy, and critical infrastructure sectors, the number of events that systems need to save and verify is increasing. The value of RTCs will also shift from "displaying time" to maintaining time continuity, recording system experiences, and providing a trusted basis for fault analysis, security, and audits.
RTC Requirements and Technological Directions in AI Systems, Source: Compiled by eefocus Research Institute
Structural Market Changes in RTC Chips
AI servers, edge computing, and smart vehicles have heightened the system's emphasis on time continuity, but it is difficult for RTCs to replicate the growth curves of GPUs, HBM, and high-speed interconnect chips. It remains a mature market with long product cycles and relatively low unit prices. New growth opportunities primarily stem from product structure upgrades, rather than every AI device adding a new standalone RTC.
Given the increasing integration of MCUs and SoCs, basic timekeeping functions are accelerating their entry into main control chips. Current mainstream MCUs typically already provide on-chip RTCs, backup power domains, calendars, alarms, and timed wake-up functions, integrating low-power management and software drivers into the development platform. For home appliances, ordinary consumer electronics, and frequently connected smart terminals, on-chip RTCs can reduce the need for standalone chips, external components, and PCB area. Devices can also be regularly calibrated via the network or mobile phones, making it a more economical solution.
This has also compressed the market space for low-end standalone RTCs. Classic general-purpose products such as the PCF8563 and DS1307 will continue to exist for a long time, primarily serving legacy platform continuations, main controllers lacking RTCs, and projects with explicit requirements for power-off timekeeping. However, basic calendar functions are no longer able to command a premium, and market competition will focus more on price, pin and register compatibility, packaging, lead time, and local services.
The true retained value of a standalone RTC lies in its isolation from the main control system. Even with an independent VBAT power domain, an on-chip RTC is still integrated with the CPU, memory, and other circuits within the same chip; once the main controller experiences a hardware failure, software crash, power anomaly, or complete replacement, the time records may also be affected. A standalone RTC has its own oscillator, counting circuit, and backup power interface, allowing it to keep time even after the main controller is completely powered off or reset, making it more suitable for fault tracing, liability determination, and long-term offline operation.
Therefore, the demand for standalone RTCs in high-reliability applications has not disappeared but has concentrated in smart metering, automotive BMS, T-Box, EDR, industrial recording equipment, medical instruments, security systems, and server management modules. These scenarios focus not only on "whether the time is correct" but also on stability in wide-temperature environments, power consumption for years of backup operation, independence after main controller failure, and whether timestamps can be verified and traced.
Power consumption and accuracy further delineate the boundaries between the two types of solutions. The backup power domain of an MCU may also maintain SRAM and wake-up logic, and its overall power consumption is typically higher than that of a dedicated RTC. Some standalone RTCs have entered the tens of nano-amps range, which can significantly extend the backup battery life in water meters, gas meters, asset trackers, and BMS in long-term storage states. Professional RTCs can also integrate crystals, temperature sensors, and compensation circuits to improve wide-temperature accuracy and long-term consistency.
The value in the RTC market is also shifting from external crystal general-purpose products to built-in crystal, TCXO, DTCXO, automotive-grade, and security-oriented products. Integrated solutions are more expensive but can reduce uncertainties caused by crystal matching, PCB parasitic parameters, and production calibration. In the future, the industry may not necessarily show rapid growth in shipment volume, but is more likely to be reflected in an increased proportion of high-value products.
Complex systems may also form a hierarchical time architecture: relying on network time or high-speed synchronization devices during normal operation, using the on-chip RTC for wake-up and scheduling during standby, and relying on the standalone RTC to maintain time continuity after power loss, failure, or network disconnection. On-chip RTCs and standalone RTCs do not have a simple substitution relationship; their market boundaries ultimately depend on the consequences of time failure. When time is only used for display, the on-chip RTC is usually sufficient; once billing, auditing, fault tracing, and security liabilities are involved, the standalone RTC remains difficult to be completely replaced.
Market Boundaries Between On-Chip RTCs and Standalone RTCs, Source: Compiled by eefocus Research Institute
RTC Competitive Landscape: Can Domestic Manufacturers Enter the High-End Market?
The circuit complexity of RTCs is not high, nor does it rely on the most advanced manufacturing processes. Compared with processors, memory, and high-speed interface chips, its accuracy, power consumption, interface, and packaging parameters are easier to compare. However, in high-reliability markets such as automotive, smart metering, industrial control, medical, and communication equipment, international manufacturers still occupy the main positions for a long time. The reason is that the competition for high-end RTCs has long gone beyond chip design itself, extending to multiple links such as crystals, packaging, calibration, mass manufacturing, quality management, and long-term supply.
In recent years, the Chinese RTC clock chip market has presented a core pattern of "foreign capital dominance and rapid catch-up by domestic brands." From 2024 to 2025, the overall market size was approximately 3.5 to 4 billion RMB, with imported brands accounting for 65-70% and domestic brands accounting for 30-35%. In 2025, the shipment volume of domestic RTC chips reached 1.86 billion units, accounting for 34.5% of the overall domestic demand, a year-on-year growth of 12.7%. It is estimated that from 2027 to 2030, the localization rate will increase to 45-50%. The global market is dominated by Japanese manufacturers (such as EPSON, etc.), which collectively hold approximately 68% of the market share.
Seiko Epson, a leading RTC chip manufacturer, represents a relatively typical vertical integration route in the RTC market. Its electronic components business covers both quartz crystals and semiconductor devices. The crystal wafers and IC silicon chips used in some active quartz products are produced by its own system, which helps to uniformly control quality, performance, capacity, and cost. This capability is particularly important for RTCs. The performance of an RTC depends not only on the IC circuit but is also affected by the initial frequency deviation of the crystal, load matching, packaging stress, environmental temperature, and long-term aging. When crystals, oscillation circuits, packaging, and calibration are completed by different suppliers, the device combination will produce more dispersion. Manufacturers with vertical integration capabilities can simultaneously optimize crystal characteristics, oscillation circuits, and temperature compensation algorithms, making it easier to control wide-temperature performance, batch variations, and long-term drift.
Comprehensive IDMs such as NXP, STMicroelectronics, Renesas, and Microchip follow a platform-based route. These manufacturers may not necessarily lead in every RTC parameter, but they possess mature automotive and industrial customer systems, as well as product portfolios of MCUs, processors, power management, and security chips. RTCs can enter customer designs along with the complete platform, and software drivers, quality certifications, technical support, and supplier management can also be completed within the same system. For large customers, this platform continuity is often more important than the price difference of a single device.
Professional manufacturers such as Maxim under ADI and Micro Crystal focus on high-precision temperature compensation, miniaturization, and special applications. TCXO, DTCXO, and built-in crystal solutions can reduce customers' work in handling crystal matching, temperature calibration, and PCB parasitic parameters, making them more suitable for industrial instruments, medical electronics, smart metering, and communication equipment. Their business logic is not to compete for the lowest-priced general-purpose market, but to obtain product premiums through accuracy, size, and reliability.
For domestic manufacturers, the harder-to-replicate advantage of international manufacturers comes from their long-term mass production track record. High-end customers rarely decide on RTC selection based solely on data sheets; they also pay attention to frequency dispersion across different batches, high and low-temperature startup, drift after reflow soldering, field failure rates, and the supplier's ability to handle anomalies. Models that have been running in a large number of automotive, instrument, and industrial equipment for many years have accumulated complete test and application data. If customers change suppliers, they would instead need to re-verify accuracy, power switching, anti-interference, and long-term reliability.
Comparison of International RTC Suppliers, Source: Compiled by eefocus Research Institute
For an RTC with a limited unit price, the procurement savings brought by replacement may not be high, but the cost of re-verification and field failure could be much greater. This gives the high-end RTC market strong customer stickiness and also explains why international manufacturers can maintain their advantages for a long time.
In recent years, the development of domestic RTCs has clearly gone beyond early compatible substitution. In the past, domestic manufacturers mainly focused on classic models such as the PCF8563, DS1302, and DS1307, entering the home appliance, security, attendance, and general industrial markets through pin compatibility, register compatibility, price, and local services. Today, domestic products have extended towards built-in crystals, temperature compensation, ultra-low power consumption, small packages, and automotive-grade directions.
The general-purpose market has become a relatively mature field for domestic RTCs. I²C and SPI interfaces and basic calendar functions are highly standardized, and domestic solutions have obvious advantages in cost, lead time, and software migration. Smart metering has become an important transitional market for domestic manufacturers to move from consumer-grade to industrial-grade. Demands such as long-term power-off retention, calendar billing, low power consumption, and anti-tampering have helped domestic enterprises accumulate more experience in packaging and testing, batch management, and reliability.
Some domestic manufacturers have launched temperature-compensated RTCs with a nominal accuracy reaching ±2ppm to ±5ppm, and have made progress in power consumption of hundreds of nano-amps, built-in crystals, and small packages. The author believes that the current status of domestic RTCs is that single-point parameters have achieved breakthroughs, but wide-temperature consistency, long-term reliability, and overall quality control still need to be improved.
Comparison of Some Domestic RTC Suppliers, Source: Compiled by eefocus Research Institute
The automotive market has further raised the threshold. AEC-Q100 certification is only the basis for entering the supply chain, followed by functional safety processes, extremely low defect rates, PPAP, batch traceability, PCN management, failure analysis, and long-term supply. OEMs usually require chip suppliers to provide supply and traceability commitments for 15 years or even longer. These capabilities require support from real projects, long-term manufacturing data, and stable operations, making it difficult to quickly establish them with just one or two generations of products.
Therefore, domestic RTCs have completed the first stage dominated by general-purpose compatibility and cost substitution, and are entering the second stage of high-precision, low-power, and automotive-grade products. The high-end barriers of international manufacturers can also be divided into three layers: the bottom layer is crystal, IC, and temperature compensation technology; the middle layer is scale manufacturing, batch consistency, and quality systems; the top layer is customer track records, long-term supply, and brand trust.
Conclusion: Where Is the Value of RTCs Reflected in the AI Era?
In the past, RTCs were often understood as basic chips that maintain date and time. The fact that time is not lost after the device is shut down and the correct date can still be displayed upon restart seemed to be its entire value. However, in AI servers, robots, smart vehicles, industrial equipment, and energy systems, the precision of time has become extremely important.
GPUs are responsible for completing calculations, high-speed clocks are responsible for maintaining data link synchronization, and network time is responsible for calibration among different nodes. The task undertaken by the RTC is quieter and more fundamental: when the main processor stops running, the device enters deep sleep, the network connection is interrupted, or the power suddenly disappears, it still preserves a continuous local timeline. Only after the system recovers can engineers know when the temperature anomaly occurred, when the power was interrupted, which software upgrade preceded the fault, and whether a certain alert appeared before or after the restart.
This is also the reason why, although RTC chips account for a small proportion in the AI value chain, their importance is increasing. The value of standalone RTCs is concentrated in high-reliability systems that cannot tolerate time loss. In the future, the focus of competition for high-end products will gradually shift from calendars, alarms, and single ppm parameters to wide-temperature stability, long-term low power consumption, independent power domains, secure timestamps, anti-tampering, and reliable operation for over ten years.
For domestic RTCs, having parameters close to international products is only the first step. What truly determines whether they can enter high-value markets such as automotive, industrial, and smart metering is whether these parameters can hold true across different batches, complex environments, and long lifecycles, and ultimately be transformed into verifiable mass production track records and customer trust.