Abstract: In stacked transistor device, such as a complementary field-effect-transistor (CFET) device, different strain materials may be used in different layers, e.g., a tensile material is deposited in a first isolation region in the PMOS layer, and a compressive material is deposited in second isolation region in the NMOS layer. The strain materials may be stacked, such that the second isolation region may be positioned over the first isolation region. In some cases, in one or both of the isolation regions, a liner material is included between the strain material and the source and drain regions. Certain embodiments provide independent tuning of strain forces in a stacked transistor device. Different materials are selected for different layers in the stacked device to provide favorable performance enhancement or tuning (e.g., adjustment of the threshold voltage) in NMOS and PMOS layers.

Abstract: Methods for doping 2D transistor devices and resulting architectures. The use and placement of oxide dopants, such as, but not limited to, GeOx, enable control over threshold voltage performance and contact resistance of 2D transistor devices. Architectures include distinct stoichiometry compositions.

Abstract: In one embodiment, an integrated circuit structure includes a first transistor device comprising a first gate stack and a second transistor device comprising a second gate stack. The second transistor device is spaced a first distance laterally from the first transistor device. The structure further includes a dielectric region between the first gate stack and the second gate stack. The dielectric region is spaced a second distance laterally from the first transistor device, where the first distance is substantially twice the second distance.

Abstract: A low strain transfer protective layer is formed on a transition metal dichalcogenide (TMD) monolayer to enable the transfer of the TMD monolayer from a growth substrate to a target substrate with little or no strain-induced damage to the TMD monolayer. Transfer of a TMD monolayer from a growth substrate to a target substrate comprises two transfers, a first transfer from the growth substrate to a carrier wafer and a second transfer from the carrier wafer to the target substrate. Transfer of the TMD monolayer from the growth substrate to the carrier wafer comprises mechanically lifting off the TMD monolayer from the growth substrate. The low strain transfer protective layer can limit the amount of strain transferred from the carrier wafer to the TMD monolayer during lift-off. The carrier wafer and protective layer are separated from the TMD monolayer after attachment of the TMD monolayer to the target substrate.

Abstract: Technologies for reducing the impact of inductors on electrical traces are disclosed. In an illustrative embodiment, conductive ink is applied in a silk screen layer on top of a solder mask of a circuit board. The conductive ink forms shield regions under and near where inductors are placed and/or where a power plane is routed. The conductive shield regions may be coupled to a ground plane in the circuit board. The conductive shield regions can partially shield traces under and near the inductor, reducing the noise induced on nearby traces. The conductive shield regions can allow traces for high-speed input/output signals to be routed closer to the inductor, reducing the size, number of layers, and/or cost of the circuit board. In some embodiments, the conductive shield regions can shield emissions from the power plane, reducing interference on antennas of a device.

Abstract: A disaggregated processor package can be configured to accept interchangeable chiplets. Interchangeability is enabled by specifying a standard physical interconnect for chiplets that can enable the chiplet to interface with a fabric or bridge interconnect. Chiplets from different IP designers can conform to the common interconnect, enabling such chiplets to be interchangeable during assembly. The fabric and bridge interconnects logic on the chiplet can then be configured to confirm with the actual interconnect layout of the on-board logic of the chiplet. Additionally, data from chiplets can be transmitted across an inter-chiplet fabric using encapsulation, such that the actual data being transferred is opaque to the fabric, further enable interchangeability of the individual chiplets. With such an interchangeable design, cache or DRAM memory can be inserted into memory chiplet slots, while compute or graphics chiplets with a higher or lower core count can be inserted into logic chiplet slots.

Abstract: Microelectronic assemblies, related devices and methods, are disclosed herein. In some embodiments, a microelectronic assembly may include a first layer having first dies in a first insulating material; a second layer on the first layer, the second layer including second dies having a first thickness and third dies having a second thickness different than the first thickness, the second dies and the third dies in a second insulating material, wherein the second dies and third dies have a first surface and an opposing second surface, and wherein the first surfaces of the second and third dies have a combined surface area between 3,000 square millimeters (mm2) and 9,000 mm2; and a redistribution layer (RDL) between the first layer and the second layer, the RDL including conductive pathways, wherein the first dies are electrically coupled to the second dies and the third dies by the conductive pathways and by interconnects.

Abstract: Quantum dot devices, and related systems and methods, are disclosed herein. In some embodiments, a quantum dot device may include a quantum well stack; a plurality of first gates above the quantum well stack; and a plurality of second gates above the quantum well stack; wherein the plurality of first gates are arranged in electrically continuous rows extending in a first direction, and the plurality of second gates are arranged in electrically continuous rows extending in a second direction perpendicular to the first direction.

Abstract: Techniques to perform time recovery from attacks on delayed authentication in a time synchronized network are described. One embodiment comprises a method for decoding time information and a message authentication code (MAC) from a time message, the time information to synchronize a local clock for a device to a network time of a time synchronized network (TSN), and the MAC to authenticate the time message, determining whether the time message is authentic using the MAC, discarding the time information when the time message is not authentic, performing a bounded search to identify authentic time information using the MAC, and passing the authentic time information to a clock manager to synchronize the local clock to the network time of the TSN when the authentic time information is identified. Other embodiments are described and claimed.

Abstract: An integrated circuit includes an update controller circuit, updatable logic circuits, and an output circuit. The update controller circuit is configured to control an output signal of the output circuit that is provided to an external conductor during reconfiguration of the updatable logic circuits.

Abstract: Systems and methods for improving cache efficiency and utilization are disclosed. In one embodiment, a graphics processor includes processing resources to perform graphics operations and a cache controller of a cache memory that is coupled to the processing resources. The cache controller is configured to set an initial aging policy using an aging field based on age of cache lines within the cache memory and to determine whether a hint or an instruction to indicate a level of aging has been received.

Abstract: Apparatuses including a graphics processing unit, graphics multiprocessor, or graphics processor having an error detection correction logic for cache memory or shared memory are disclosed. In one embodiment, a graphics multiprocessor includes cache or local memory for storing data and error detection correction circuitry integrated with or coupled to the cache or local memory. The error detection correction circuitry is configured to perform a tag read for data of the cache or local memory to check error detection correction information.

Abstract: Technologies for secure device configuration and management include a computing device having an I/O device. A trusted agent of the computing device is trusted by a virtual machine monitor of the computing device. The trusted agent securely commands the I/O device to enter a trusted I/O mode, securely commands the I/O device to set a global lock on configuration registers, receives configuration data from the I/O device, and provides the configuration data to a trusted execution environment. In the trusted I/O mode, the I/O device rejects a configuration command if a configuration register associated with the configuration command is locked and the configuration command is not received from the trusted agent. The trusted agent may provide attestation information to the trusted execution environment. The trusted execution environment may verify the configuration data and the attestation information. Other embodiments are described and claimed.

Abstract: Systems and methods for improving cache efficiency and utilization are disclosed. In one embodiment, a graphics processor includes processing resources to perform graphics operations and a cache controller of a cache coupled to the processing resources. The cache controller is configured to control cache priority by determining whether default settings or an instruction will control cache operations for the cache.

Abstract: Techniques for quality of service (QoS) support for input/output devices and other agents are described. In embodiments, a processing device includes execution circuitry to execute a plurality of software threads; hardware to control monitoring or allocating, among the plurality of software threads, one or more shared resources; and configuration storage to enable the monitoring or allocating of the one or more shared resources among the plurality of software threads and one or more channels through which one or more devices are to be connected to the one or more shared resources.

Abstract: Embodiments are generally directed to cache structure and utilization. An embodiment of an apparatus includes one or more processors including a graphics processor; a memory for storage of data for processing by the one or more processors; and a cache to cache data from the memory; wherein the apparatus is to provide for dynamic overfetching of cache lines for the cache, including receiving a read request and accessing the cache for the requested data, and upon a miss in the cache, overfetching data from memory or a higher level cache in addition to fetching the requested data, wherein the overfetching of data is based at least in part on a current overfetch boundary, and provides for data is to be prefetched extending to the current overfetch boundary.

Abstract: In one embodiment, a device includes a fiber array unit (FAU) coupled to a photonics integrated circuit (PIC) die. The PIC die includes a cavity defined at an edge of the PIC die, with outer edges of the cavity being formed at an angle less than 90 degrees with respect to a bottom surface of the cavity. The PIC die further includes first waveguides protruding into the cavity of the PIC die. The FAU includes a shelf portion extending from a body portion, and a plurality of second waveguides protruding from an outer edge of the shelf portion opposite the body portion. The FAU further includes alignment structures on outer edges of the shelf portion that are in contact with the angled edges of the cavity of the PIC die.

Abstract: Embodiments may comprise N-path filter circuitry with tunable radio frequency selectivity and up to 80 decibels per decade roll-off. The N-path filter may comprise at least one input transistor, wherein the at least one input transistor comprises a channel and a gate. A first end of the channel is coupled with a receiver circuitry input, wherein a second end of the channel is coupled with a load. The gate of the at least one input transistor is coupled with a clock circuitry input. The load may comprise a fourth order, all-pole driving point impedance. The impedance may shunt the second end of the channel to a circuit ground or a low voltage circuit rail via the impedance. And the impedance may comprise a first active impedance circuit coupled in series with a second active impedance circuit.

Abstract: Embodiments described herein provide an apparatus comprising a plurality of processing resources including a first processing resource and a second processing resource, a memory communicatively coupled to the first processing resource and the second processing resource, and a processor to receive data dependencies for one or more tasks comprising one or more producer tasks executing on the first processing resource and one or more consumer tasks executing on the second processing resource and move a data output from one or more producer tasks executing on the first processing resource to a cache memory communicatively coupled to the second processing resource. Other embodiments may be described and claimed.

Abstract: An integrated circuit includes a communication controller circuit for exchanging communications with a device external to the integrated circuit through a signal line, a current circuit coupled to the signal line, and a current controller circuit for causing the current circuit to provide a constant current to the signal line while a signal is transmitted through the signal line based on a command generated by the communication controller circuit.

Abstract: Technologies for providing secure utilization of tenant keys include a compute device. The compute device includes circuitry configured to obtain a tenant key. The circuitry is also configured to receive encrypted data associated with a tenant. The encrypted data defines an encrypted image that is executable by the compute device to perform a workload on behalf of the tenant in a virtualized environment. Further, the circuitry is configured to utilize the tenant key to decrypt the encrypted data and execute the workload without exposing the tenant key to a memory that is accessible to another workload associated with another tenant.

Abstract: An apparatus includes a host interface, a network interface, and programmable circuitry communicably coupled to the host interface and the network interface, the programmable circuitry comprising one or more processors are to implement network interface functionality and are to receive a prompt directed to an artificial intelligence (AI) model hosted by a host device communicably coupled to the host interface, apply a prompt tuning model to the prompt to generate an initial augmented prompt, compare the initial augmented prompt for a match with stored data of a prompt augmentation tracking table comprising real-time datacenter trend data and cross-network historical augmentation data from programmable network interface devices in a datacenter hosting the apparatus, generate, in response to identification of the match with the stored data, a final augmented prompt based on the match, and transmit the final augmented prompt to the AI model.

Abstract: Embodiments described herein include software, firmware, and hardware logic that provides techniques to perform arithmetic on sparse data via a systolic processing unit. One embodiment provides for data aware sparsity via compressed bitstreams. One embodiment provides for block sparse dot product instructions. One embodiment provides for a depth-wise adapter for a systolic array.

Abstract: Glass cores including protruding through glass vias and related methods are disclosed herein. An example substrate disclosed herein includes a glass core including a surface and a copper through glass via (TGV) extending through the glass core, the TGV including a protrusion extending from the surface.

Abstract: An IC device may include memory layers over a logic layer. A memory layer may include memory arrays and one or more peripheral circuits coupled to the memory arrays. A memory array may include memory cells arranged in rows and columns. A row of memory cells may be associated with a word line. A column of memory cells may be associated with a bit line. The logic layer includes one or more logic circuits that can control data read operations and data write operations of the memory layers. The logic layer may also include a power interconnect, which facilitates power delivery to the memory layers, and a signal interconnect, which facilitates signal transmission within the IC device. The IC device may further include vias that couple the memory layers to the logic layer. Each via may be connected to one or more memory layers and the logic layer.

Abstract: An IC device may include memory layers over a logic layer. A memory layer includes memory arrays, each of which includes memory cells arranged in rows and columns. A row of memory cells may be associated with a word line. A column of memory cells may be associated with a bit line. Bit lines of different memory arrays may be coupled using one or more vias or source/drain electrodes of transistors in the memory arrays. Alternatively, word lines of different memory arrays may be coupled using one or more vias or gate electrodes of transistors in the memory arrays. The logic layer has a logic circuit that can control data read operations and data write operations of the memory layers. The logic layer may include a power interconnect, which facilitates power delivery to the memory layers, and a signal interconnect, which facilitates signal transmission within the memory device.

Abstract: An IC device may include memory layers bonded to a logic layer with inclination. An angle between a memory layer and the logic layer may be in a range from approximately 0 to approximately 90 degrees. The memory layers may be over the logic layer. The IC device may include one or more additional logic layers that are parallel to a memory layer or perpendicular to a memory layer. The one or more additional logic layers may be over the logic layer. A memory layer may include memory cells. The logic layer may include logic circuits (e.g., sense amplifier, word line driver, etc.) that control the memory cells. Bit lines (or word lines) in different memory layers may be coupled to each other. A bit line and a word line in a memory layer may be controlled by logic circuits in different logic layers.

Abstract: Methods of selectively transferring portions of layers between substrates, and devices and systems formed using the same, are disclosed herein. In one embodiment, a first substrate with a release layer and a layer of integrated circuit (IC) components over the release layer is received, and a second substrate with one or more adhesive areas is received. The release layer on the first substrate is weakened. The first substrate is partially bonded to the second substrate, such that a subset of IC components on the first substrate are bonded to the adhesive areas on the second substrate. The first substrate is then separated from the second substrate, and the subset of IC components bonded to the second substrate are separated from the first substrate and remain on the second substrate.

Abstract: Methods of selectively transferring portions of layers between substrates, and devices and systems formed using the same, are disclosed herein. In one embodiment, a first substrate with a layer of integrated circuit (IC) components is received, and a second substrate with one or more adhesive areas is received. The first substrate is partially bonded to the second substrate, such that a subset of IC components on the first substrate are bonded to the adhesive areas on the second substrate. The first substrate is then separated from the second substrate, and the subset of IC components bonded to the second substrate are separated from the first substrate and remain on the second substrate.

Abstract: Methods of selectively transferring integrated circuit (IC) components between substrates, and devices and systems formed using the same, are disclosed herein. In one embodiment, a first substrate with a release layer and a layer of IC components over the release layer is received, and a second substrate with one or more adhesive areas is received. The layer of IC components may include one or more waveguides, ring resonators, drivers, photodetectors, transimpedance amplifiers, and/or electronic integrated circuits. The first substrate is partially bonded to the second substrate, such that a subset of IC components on the first substrate are bonded to the adhesive areas on the second substrate. The first substrate is then separated from the second substrate, and the subset of IC components bonded to the second substrate are separated from the first substrate and remain on the second substrate.

Abstract: Example systems, apparatus, articles of manufacture, and methods that perform memory preservation to improve system reliability are disclosed. Example apparatus disclosed herein increment an error count after detection of an error associated with a memory cell. Example apparatus also isolate a system memory address of the memory cell based on the error count.

Abstract: Techniques are provided herein to form an integrated circuit having dielectric material formed in cavities beneath source or drain regions. The cavities may be formed within subfin portions of semiconductor devices. In one such example, a FET (field effect transistor) includes a gate structure extending around a fin or any number of nanowires of semiconductor material. The semiconductor material may extend in a first direction between source and drain regions while the gate structure extends over the semiconductor material in a second direction substantially orthogonal to the first direction. A dielectric fill may be formed in a recess beneath the source or drain regions, or a dielectric liner may be formed on sidewalls of the recess, to prevent epitaxial growth of the source or drain regions from the subfins. Removal of the semiconductor subfin from the backside may then be performed without causing damage to the source or drain regions.

Abstract: One embodiment provides a graphics processor comprising an interface to a system interconnect and a graphics processor coupled to the interface, the graphics processor comprising circuitry configured to compact sample data for multiple sample locations of a pixel, map the multiple sample locations to memory locations that store compacted sample data, the memory locations in a memory of the graphics processor, apply lossless compression to the compacted sample data, and update a compression control surface associated with the memory locations, the compression control surface to specify a compression status for the memory locations

Abstract: An IC device may include one or more vias for delivering power to one or more transistors in the IC device. A via may have one or more widened ends to increase capacitance and decrease resistance. A transistor may include a source electrode over a source region and a drain electrode over a drain region. The source region or drain region may be in a support structure that has one or more semiconductor materials. The via has a body section and two end sections, the body section is between the end sections. One or both end sections are wider than the body section, e.g., by approximately 6 nanometers to approximately 12 nanometers. One end section is connected to an interconnect at the backside of the support structure. The other end section is connected to a jumper, which is connected to the source electrode or drain electrode.

Abstract: An example IC structure includes a first layer comprising a plurality of transistors; a second layer comprising a stack of layers of one or more insulator materials and conductive interconnect structures extending through the one or more insulator materials; a third layer comprising bonding pads, wherein the second layer is between the first layer and the third layer; and a via continuously extending between one of the bonding pads and one of the conductive interconnect structures in a bottom layer of the stack of layers or a conductive structure in the first layer, wherein the bottom layer is a layer of the stack of layers that is closer to the first layer than all other layers of the stack of layers.

Abstract: Hybrid bonding interconnect (HBI) architectures for scalability. Embodiments implement a bonding layer on a semiconductor die that includes a thick oxide layer overlaid with a thin layer of a hermetic material including silicon and at least one of carbon and nitrogen. The conductive bonds of the semiconductor die are placed in the thick oxide layer and exposed at the surface of the hermetic material. Some embodiments implement a non-bonding moisture seal ring (MSR) structure.

Abstract: Disclosed herein are microelectronic assemblies and related devices and methods. In some embodiments, a microelectronic assembly may include a glass layer having a surface, the glass layer including conductive through-glass vias (TGVs); a dielectric layer at the surface of the glass layer, the dielectric layer including conductive pathways; and interconnects between the surface of the glass layer and the dielectric layer, wherein individual interconnects electrically couple individual TGVs to individual conductive pathways. In some embodiments, the interconnects include solder or liquid metal ink. In some embodiments, the interconnects include metal-metal bonds and dielectric-dielectric bonds.

Abstract: Techniques are provided to form an integrated circuit having an airgap spacer between at least a transistor gate structure and an adjacent source or drain contact. In one such example, a FET (field effect transistor) includes a gate structure that extends around a fin or any number of nanowires (or nanoribbons or nanosheets, as the case may be) of semiconductor material. The semiconductor material may extend in a first direction between source and drain regions while the gate structure extends over the semiconductor material in a second direction. Airgaps are provided in the regions between the gate structures and the adjacent source/drain contacts. The airgaps have a low dielectric constant (e.g., around 1.0) to reduce the parasitic capacitance between the conductive structures.

Abstract: Microelectronic assemblies, related devices and methods, are disclosed herein. In some embodiments, a microelectronic assembly may include a first layer including first dies in a first insulating material; a second layer on the first layer, the second layer including second dies and third dies in a second insulating material, the second dies having a first thickness, the third dies having a second thickness different than the first thickness, and the second dies and the third dies having a surface, wherein the surfaces of the second and third dies have a combined surface area between 3,000 square millimeters (mm2) and 9,000 mm2; and a redistribution layer (RDL) between the first layer and the second layer, the RDL including conductive pathways through the RDL, wherein the first dies are electrically coupled to the second dies and the third dies by the conductive pathways through the RDL and by interconnects.

Abstract: Techniques are provided to form an integrated circuit having a gate electrode that includes at least one layer containing molybdenum. A transistor includes a gate structure having a gate electrode on a gate dielectric. The gate structure extends around a fin or any number of nanowires (or nanoribbons or nanosheets) of semiconductor material. The gate electrode includes one or more conductive layers on the gate dielectric with at least one of those conductive layers containing molybdenum (e.g., molybdenum nitride). The conductive layer having molybdenum may be used during the formation of the gate dielectric (e.g., during an annealing process), thus resulting in a higher quality gate dielectric.

Abstract: Technologies for tunable lasers in a photonic integrated circuit (PIC) die are disclosed. In an illustrative embodiment, a lidar system includes a PIC die with two lasers. The PIC die includes a switch to switch between the output of the first laser and the output of the second laser. Each laser can be tuned to different peaks of a Bragg grating in the cavity of the laser, and each laser can be frequency swept within the peak of the Bragg grating. In operation, one laser is changed to a different peak of the Bragg grating and allowed to stabilize while the other laser is selected for output and frequency swept. In this manner, one laser stabilizes while the other one is used. Such a lidar system can implement frequency-modulated continuous-wave (FMCW) lidar with a stable, compact laser source.

Abstract: An IC device may include functional regions as well as replica cells and filler cells that can reduce local layout effect in the IC device. A functional region includes functional cells, e.g., logic cell or memory cells. A white space may be between a first functional region and a second functional region. A first portion of the white space may be filled with replica cells, each of which is a replica of a cell in the first functional region. A second portion of the white space may be filled with filler cells that are not functional. The first function region is closer to the replica cells than to the filler cells. A third portion of the white space may be filled with replica cells, each of which is a replica of a cell in the second functional region. The second portion is between the first portion and the third portion.

Abstract: Apparatuses, methods, and computer readable media for regulating command submission to a shared device. A processor may receive a command for an operation to be performed by another device. The processor may determine an identifier of an address space of a process associated with the command. The processor may determine whether to accept or reject the command.

Abstract: Technologies for fiber array unit (FAU) lid designs are disclosed. In one embodiment, channels in the lid allow for suction to be applied to fibers that the lid covers, pulling the fibers into place in a V-groove. The suction can hold the fibers in place as the fiber array unit is mated with a photonic integrated circuit (PIC) die. Additionally or alternatively, channels can be on pitch, allowing for pulling the FAU towards a PIC die as well as sensing the position and alignment of the FAU to the PIC die. In another embodiment, a warpage amount of a PIC die is characterized, and a FAU lid with a similar warpage is fabricated, allowing for the FAU to position fibers correctly relative to waveguides in the PIC die. In another embodiment, a FAU has an extended lid, which can provide fiber protection as well as position and parallelism tolerance control.

Abstract: Embodiments described herein include software, firmware, and hardware logic that provides techniques to perform arithmetic on sparse data via a systolic processing unit. One embodiment provides techniques to optimize training and inference on a systolic array when using sparse data. One embodiment provides techniques to use decompression information when performing sparse compute operations. One embodiment enables the disaggregation of special function compute arrays via a shared reg file. One embodiment enables packed data compress and expand operations on a GPGPU. One embodiment provides techniques to exploit block sparsity within the cache hierarchy of a GPGPU.

Abstract: A microelectronic assembly includes a bridge die embedded in a substrate. The substrate includes a doped dielectric material in a layer or region directly below the bridge die, and in a layer near an upper face of the bridge die. A cavity is formed in the upper layer of the doped dielectric material for embedding the bridge die, exposing the lower layer of the doped dielectric material. After cavity formation, a selective metallization of the lower and upper layers of the doped dielectric material is performed, providing well-aligned metal layers in the region of the bridge die and the region around the bridge die.

Abstract: IC structures with air gap insulation in place of gate spacers are disclosed. An example IC structure includes a transistor comprising a channel region and a source or drain (S/D) region, a gate structure coupled to the channel region and comprising a gate electrode material and a first electrically conductive material, a S/D contact structure coupled to the S/D region and comprising a second electrically conductive material, a gap between the gate structure and the S/D contact structure, and a liner material on at least a portion of a sidewall of the gap, the liner material comprising aluminum and oxygen.

Abstract: A microelectronic assembly includes an embedded bridge die and a glass structure, such as glass patch, under the bridge die. The bridge die and the glass structure are embedded in a substrate. The assembly may further include two or more dies arranged over the substrate and coupled to the bridge die. The glass structure may include through-glass vias, and vias in the substrate below the glass structure are self-aligned to the through-glass vias. The glass structure may include an embedded passive device, such as an embedded inductor or capacitor.

Abstract: Described herein is a technique in which a plurality of distortion meshes compensate for radial and chromatic aberrations created by optical lenses. The plurality of distortion meshes may include different lens specific parameters that allow the distortion meshes to compensate for chromatic aberrations created within received images. The plurality of distortion meshes may correspond to a red color channel, green color channel, or blue color channel to compensate for the chromatic aberrations. The distortion meshes may also include shaped distortions and grids to compensate for radial distortions, such as pin cushion distortions. In one example, the system uses a barrel-shaped distortion and a triangulation grid to compensate for the distortions created when the received image is displayed on a lens.

Abstract: Examples include techniques for artificial intelligence (AI) capabilities at a network switch. These examples include receiving a request to register a neural network for loading to an inference resource located at the network switch and loading the neural network based on information included in the request to support an AI service to be provided by users requesting the AI service.