diff --git a/src/plugins/ucx/meson.build b/src/plugins/ucx/meson.build
index aebfa92f9..2bf1ff038 100644
--- a/src/plugins/ucx/meson.build
+++ b/src/plugins/ucx/meson.build
@@ -16,26 +16,20 @@
 ucx_utils_dep = declare_dependency(link_with: ucx_utils_lib, include_directories: utils_inc_dirs )
 asio_dep = [dependency('asio', required: true)]
 
-compile_flags = []
-if cuda_dep.found()
-    compile_flags = [ '-DHAVE_CUDA' ]
-endif
-
 if 'UCX' in static_plugins
     ucx_backend_lib = static_library('UCX',
                'ucx_backend.cpp', 'ucx_backend.h', 'ucx_plugin.cpp',
-               dependencies: [nixl_infra, ucx_utils_dep, serdes_interface, cuda_dep, ucx_dep, thread_dep, nixl_common_dep, asio_dep],
+               dependencies: [nixl_infra, ucx_utils_dep, serdes_interface, ucx_dep, thread_dep, nixl_common_dep, asio_dep],
                include_directories: nixl_inc_dirs,
                install: false,
-               cpp_args : compile_flags,
                name_prefix: 'libplugin_')  # Custom prefix for plugin libraries
 else
     ucx_backend_lib = shared_library('UCX',
                'ucx_backend.cpp', 'ucx_backend.h', 'ucx_plugin.cpp',
-               dependencies: [nixl_infra, ucx_utils_dep, serdes_interface, cuda_dep, ucx_dep, thread_dep, nixl_common_dep, asio_dep],
+               dependencies: [nixl_infra, ucx_utils_dep, serdes_interface, ucx_dep, thread_dep, nixl_common_dep, asio_dep],
                include_directories: nixl_inc_dirs,
                install: true,
-               cpp_args : compile_flags + ['-fPIC'],
+               cpp_args : ['-fPIC'],
                name_prefix: 'libplugin_',  # Custom prefix for plugin libraries
                install_dir: plugin_install_dir,
                install_rpath: '$ORIGIN/..')
diff --git a/src/plugins/ucx/ucx_backend.cpp b/src/plugins/ucx/ucx_backend.cpp
index 9be55fb28..56af4f0b5 100644
--- a/src/plugins/ucx/ucx_backend.cpp
+++ b/src/plugins/ucx/ucx_backend.cpp
@@ -30,14 +30,6 @@
 #include "absl/strings/numbers.h"
 #include "absl/strings/str_join.h"
 #include <asio.hpp>
-
-#ifdef HAVE_CUDA
-
-#include <cuda.h>
-#include <cuda_runtime.h>
-
-#endif
-
 namespace {
     void moveNotifList(notif_list_t &src, notif_list_t &tgt)
     {
@@ -48,223 +40,6 @@ namespace {
     }
 }
 
-/****************************************
- * CUDA related code
- *****************************************/
-
-class nixlUcxCudaCtx {
-public:
-#ifdef HAVE_CUDA
-    CUcontext pthrCudaCtx;
-    int myDevId;
-
-    nixlUcxCudaCtx() {
-        pthrCudaCtx = NULL;
-        myDevId = -1;
-    }
-#endif
-    void cudaResetCtxPtr();
-    int cudaUpdateCtxPtr(void *address, int expected_dev, bool &was_updated);
-    int cudaSetCtx();
-};
-
-class nixlUcxCudaDevicePrimaryCtx {
-#ifndef HAVE_CUDA
-public:
-    bool push() { return false; }
-    void pop() {};
-#else
-    static constexpr int defaultCudaDeviceOrdinal = 0;
-    int m_ordinal{defaultCudaDeviceOrdinal};
-    CUdevice m_device{CU_DEVICE_INVALID};
-    CUcontext m_context{nullptr};
-public:
-
-    bool push() {
-        CUcontext context;
-
-        const auto res = cuCtxGetCurrent(&context);
-        if (res != CUDA_SUCCESS || context != nullptr) {
-            return false;
-        }
-
-        if (m_context == nullptr) {
-            CUresult res = cuDeviceGet(&m_device, m_ordinal);
-            if (res != CUDA_SUCCESS) {
-                return false;
-            }
-
-            res = cuDevicePrimaryCtxRetain(&m_context, m_device);
-            if (res != CUDA_SUCCESS) {
-                m_context = nullptr;
-                return false;
-            }
-        }
-
-        return cuCtxPushCurrent(m_context) == CUDA_SUCCESS;
-    }
-
-    void pop() {
-        cuCtxPopCurrent(nullptr);
-    }
-
-    ~nixlUcxCudaDevicePrimaryCtx() {
-        if (m_context != nullptr) {
-            cuDevicePrimaryCtxRelease(m_device);
-        }
-    }
-#endif
-};
-
-class nixlUcxCudaCtxGuard {
-    nixlUcxCudaDevicePrimaryCtxPtr m_primary;
-public:
-    nixlUcxCudaCtxGuard(nixl_mem_t nixl_mem,
-                        nixlUcxCudaDevicePrimaryCtxPtr primary) {
-        if (nixl_mem == VRAM_SEG && primary && primary->push()) {
-            m_primary = primary;
-        }
-    }
-    ~nixlUcxCudaCtxGuard() {
-        if (m_primary) {
-            m_primary->pop();
-        }
-    }
-};
-
-#ifdef HAVE_CUDA
-
-static int cudaQueryAddr(void *address, bool &is_dev,
-                         CUdevice &dev, CUcontext &ctx)
-{
-    CUmemorytype mem_type = CU_MEMORYTYPE_HOST;
-    uint32_t is_managed = 0;
-#define NUM_ATTRS 4
-    CUpointer_attribute attr_type[NUM_ATTRS];
-    void *attr_data[NUM_ATTRS];
-    CUresult result;
-
-    attr_type[0] = CU_POINTER_ATTRIBUTE_MEMORY_TYPE;
-    attr_data[0] = &mem_type;
-    attr_type[1] = CU_POINTER_ATTRIBUTE_IS_MANAGED;
-    attr_data[1] = &is_managed;
-    attr_type[2] = CU_POINTER_ATTRIBUTE_DEVICE_ORDINAL;
-    attr_data[2] = &dev;
-    attr_type[3] = CU_POINTER_ATTRIBUTE_CONTEXT;
-    attr_data[3] = &ctx;
-
-    result = cuPointerGetAttributes(4, attr_type, attr_data, (CUdeviceptr)address);
-
-    is_dev = (mem_type == CU_MEMORYTYPE_DEVICE);
-
-    return (CUDA_SUCCESS != result);
-}
-
-// This routine finds the CUDA context matching for the given input address.
-int nixlUcxCudaCtx::cudaUpdateCtxPtr(void *address, int expected_dev, bool &was_updated)
-{
-    bool is_dev;
-    CUdevice dev;
-    CUcontext ctx;
-    int ret;
-
-    was_updated = false;
-
-    /* TODO: proper error codes and log outputs through this method */
-    if (expected_dev == -1) {
-        return -1;
-    }
-
-    ret = cudaQueryAddr(address, is_dev, dev, ctx);
-    if (ret) {
-        return ret;
-    }
-
-    if (!is_dev) {
-        return 0;
-    }
-
-    if (dev != expected_dev) {
-        // User provided address that does not match dev_id
-        return -1;
-    }
-
-    pthrCudaCtx = ctx;
-    was_updated = true;
-    myDevId = expected_dev;
-
-    return 0;
-}
-
-int nixlUcxCudaCtx::cudaSetCtx()
-{
-    CUresult result;
-    if (NULL == pthrCudaCtx) {
-        return 0;
-    }
-
-    result = cuCtxSetCurrent(pthrCudaCtx);
-
-    return (CUDA_SUCCESS == result);
-}
-
-#else
-
-int nixlUcxCudaCtx::cudaUpdateCtxPtr(void *address, int expected_dev, bool &was_updated)
-{
-    was_updated = false;
-    return 0;
-}
-
-int nixlUcxCudaCtx::cudaSetCtx() {
-    return 0;
-}
-
-#endif
-
-
-void nixlUcxEngine::vramInitCtx()
-{
-    cudaCtx = std::make_unique<nixlUcxCudaCtx>();
-}
-
-int
-nixlUcxEngine::vramUpdateCtx(void *address, uint64_t dev_id, bool &restart_reqd) {
-    int ret;
-    bool was_updated;
-
-    restart_reqd = false;
-
-    if(!cuda_addr_wa) {
-        // Nothing to do
-        return 0;
-    }
-
-    ret = cudaCtx->cudaUpdateCtxPtr(address, dev_id, was_updated);
-    if (ret) {
-        return ret;
-    }
-
-    restart_reqd = was_updated;
-
-    return 0;
-}
-
-int nixlUcxEngine::vramApplyCtx()
-{
-    if(!cuda_addr_wa) {
-        // Nothing to do
-        return 0;
-    }
-
-    return cudaCtx->cudaSetCtx();
-}
-
-void nixlUcxEngine::vramFiniCtx()
-{
-    cudaCtx.reset();
-}
-
 /****************************************
  * UCX request management
 *****************************************/
@@ -429,9 +204,7 @@ class nixlUcxBackendH : public nixlBackendReqH {
  */
 class nixlUcxThread {
 public:
-    nixlUcxThread(const nixlUcxEngine *engine, std::function<void()> init, size_t num_workers)
-        : engine_(engine),
-          init_(std::move(init)) {
+    nixlUcxThread(const nixlUcxEngine *engine, size_t num_workers) : engine_(engine) {
         workers_.reserve(num_workers);
     }
 
@@ -477,7 +250,6 @@ class nixlUcxThread {
     void
     operator()() {
         tlsThread() = this;
-        init_();
         threadActive_->set_value();
         run();
     }
@@ -506,7 +278,6 @@ class nixlUcxThread {
 
 private:
     const nixlUcxEngine *engine_;
-    std::function<void()> init_;
     std::vector<nixlUcxWorker *> workers_;
     std::vector<size_t> workerIds_;
     std::unique_ptr<std::thread> thread_;
@@ -515,11 +286,8 @@ class nixlUcxThread {
 
 class nixlUcxSharedThread : public nixlUcxThread {
 public:
-    nixlUcxSharedThread(const nixlUcxEngine *engine,
-                        std::function<void()> init,
-                        size_t num_workers,
-                        nixlTime::us_t delay)
-        : nixlUcxThread(engine, std::move(init), num_workers) {
+    nixlUcxSharedThread(const nixlUcxEngine *engine, size_t num_workers, nixlTime::us_t delay)
+        : nixlUcxThread(engine, num_workers) {
         if (pipe(controlPipe_) < 0) {
             throw std::runtime_error("Couldn't create progress thread control pipe");
         }
@@ -607,8 +375,7 @@ nixlUcxThreadEngine::nixlUcxThreadEngine(const nixlBackendInitParams &init_param
     }
 
     size_t num_workers = getWorkers().size();
-    thread_ = std::make_unique<nixlUcxSharedThread>(
-        this, [this]() { nixlUcxEngine::vramApplyCtx(); }, num_workers, init_params.pthrDelay);
+    thread_ = std::make_unique<nixlUcxSharedThread>(this, num_workers, init_params.pthrDelay);
     for (size_t i = 0; i < num_workers; i++) {
         thread_->addWorker(getWorkers()[i].get(), i);
     }
@@ -619,13 +386,6 @@ nixlUcxThreadEngine::~nixlUcxThreadEngine() {
     thread_->join();
 }
 
-int
-nixlUcxThreadEngine::vramApplyCtx() {
-    thread_->join();
-    thread_->start();
-    return nixlUcxEngine::vramApplyCtx();
-}
-
 void
 nixlUcxThreadEngine::appendNotif(std::string remote_name, std::string msg) {
     if (nixlUcxThread::isProgressThread(this)) {
@@ -827,8 +587,8 @@ class nixlUcxCompositeBackendH : public nixlUcxBackendH {
 
 class nixlUcxDedicatedThread : public nixlUcxThread {
 public:
-    nixlUcxDedicatedThread(nixlUcxEngine *engine, std::function<void()> init, asio::io_context &io)
-        : nixlUcxThread(engine, std::move(init), 1),
+    nixlUcxDedicatedThread(nixlUcxEngine *engine, asio::io_context &io)
+        : nixlUcxThread(engine, 1),
           io_(io) {}
 
     static nixlUcxDedicatedThread *
@@ -898,11 +658,9 @@ nixlUcxThreadPoolEngine::nixlUcxThreadPoolEngine(const nixlBackendInitParams &in
 
     splitBatchSize_ = nixl_b_params_get(init_params.customParams, "split_batch_size", 1024);
 
-    auto init = [this]() { nixlUcxEngine::vramApplyCtx(); };
-
     if (init_params.enableProgTh) {
-        sharedThread_ = std::make_unique<nixlUcxSharedThread>(
-            this, init, numSharedWorkers_, init_params.pthrDelay);
+        sharedThread_ =
+            std::make_unique<nixlUcxSharedThread>(this, numSharedWorkers_, init_params.pthrDelay);
         for (size_t i = 0; i < numSharedWorkers_; i++) {
             sharedThread_->addWorker(getWorkers()[i].get(), i);
         }
@@ -914,8 +672,7 @@ nixlUcxThreadPoolEngine::nixlUcxThreadPoolEngine(const nixlBackendInitParams &in
         dedicatedThreads_.reserve(num_threads);
         for (size_t i = 0; i < num_threads; ++i) {
             size_t worker_id = numSharedWorkers_ + i;
-            dedicatedThreads_.emplace_back(
-                std::make_unique<nixlUcxDedicatedThread>(this, init, *io_));
+            dedicatedThreads_.emplace_back(std::make_unique<nixlUcxDedicatedThread>(this, *io_));
             dedicatedThreads_.back()->addWorker(getWorker(worker_id).get(), worker_id);
             dedicatedThreads_.back()->start();
         }
@@ -1014,25 +771,6 @@ nixlUcxThreadPoolEngine::sendXferRange(const nixl_xfer_op_t &operation,
     return status.load();
 }
 
-int
-nixlUcxThreadPoolEngine::vramApplyCtx() {
-    if (sharedThread_) {
-        sharedThread_->join();
-        sharedThread_->start();
-    }
-    if (io_) {
-        io_->stop();
-        for (auto &thread : dedicatedThreads_) {
-            thread->join();
-        }
-        io_->restart();
-        for (auto &thread : dedicatedThreads_) {
-            thread->start();
-        }
-    }
-    return nixlUcxEngine::vramApplyCtx();
-}
-
 void
 nixlUcxThreadPoolEngine::appendNotif(std::string remote_name, std::string msg) {
     if (nixlUcxThread::isProgressThread(this)) {
@@ -1111,17 +849,6 @@ nixlUcxEngine::nixlUcxEngine(const nixlBackendInitParams &init_params)
     auto &uw = uws.front();
     workerAddr = uw->epAddr();
     uw->regAmCallback(NOTIF_STR, notifAmCb, this);
-
-    // Temp fixup
-    if (getenv("NIXL_DISABLE_CUDA_ADDR_WA")) {
-        NIXL_INFO << "disabling CUDA address workaround";
-        cuda_addr_wa = false;
-    } else {
-        cuda_addr_wa = true;
-    }
-
-    m_cudaPrimaryCtx = std::make_shared<nixlUcxCudaDevicePrimaryCtx>();
-    vramInitCtx();
 }
 
 nixl_mem_list_t nixlUcxEngine::getSupportedMems () const {
@@ -1139,7 +866,6 @@ tlsSharedWorkerMap() {
 
 // Through parent destructor the unregister will be called.
 nixlUcxEngine::~nixlUcxEngine() {
-    vramFiniCtx();
     tlsSharedWorkerMap().erase(this);
 }
 
@@ -1218,17 +944,6 @@ nixl_status_t nixlUcxEngine::registerMem (const nixlBlobDesc &mem,
 {
     auto priv = std::make_unique<nixlUcxPrivateMetadata>();
 
-    if (nixl_mem == VRAM_SEG) {
-        bool need_restart;
-        if (vramUpdateCtx((void*)mem.addr, mem.devId, need_restart)) {
-            return NIXL_ERR_NOT_SUPPORTED;
-            //TODO Add to logging
-        }
-        if (need_restart) {
-            vramApplyCtx();
-        }
-    }
-
     // TODO: Add nixl_mem check?
     const int ret = uc->memReg((void*) mem.addr, mem.len, priv->mem, nixl_mem);
     if (ret) {
@@ -1307,8 +1022,6 @@ nixl_status_t nixlUcxEngine::loadRemoteMD (const nixlBlobDesc &input,
                                            const std::string &remote_agent,
                                            nixlBackendMD* &output)
 {
-    // Set CUDA context of first device, UCX will anyways detect proper device when sending
-    nixlUcxCudaCtxGuard guard(nixl_mem, m_cudaPrimaryCtx);
     return internalMDHelper(input.metaInfo, remote_agent, output);
 }
 
diff --git a/src/plugins/ucx/ucx_backend.h b/src/plugins/ucx/ucx_backend.h
index bf9e6546d..907222ee9 100644
--- a/src/plugins/ucx/ucx_backend.h
+++ b/src/plugins/ucx/ucx_backend.h
@@ -93,16 +93,6 @@ class nixlUcxPublicMetadata : public nixlBackendMD {
     std::vector<std::unique_ptr<nixl::ucx::rkey>> rkeys_;
 };
 
-// Forward declaration of CUDA context
-// It is only visible in ucx_backend.cpp to ensure that
-// HAVE_CUDA works properly
-// Once we will introduce static config (i.e. config.h) that
-// will be part of NIXL installation - we can have
-// HAVE_CUDA in h-files
-class nixlUcxCudaCtx;
-class nixlUcxCudaDevicePrimaryCtx;
-using nixlUcxCudaDevicePrimaryCtxPtr = std::shared_ptr<nixlUcxCudaDevicePrimaryCtx>;
-
 class nixlUcxEngine : public nixlBackendEngine {
 public:
     static std::unique_ptr<nixlUcxEngine>
@@ -247,9 +237,6 @@ class nixlUcxEngine : public nixlBackendEngine {
     void
     getNotifsImpl(notif_list_t &notif_list);
 
-    virtual int
-    vramApplyCtx();
-
     virtual void
     appendNotif(std::string remote_name, std::string msg);
 
@@ -265,13 +252,6 @@ class nixlUcxEngine : public nixlBackendEngine {
     nixlUcxEngine(const nixlBackendInitParams &init_params);
 
 private:
-    void
-    vramInitCtx();
-    void
-    vramFiniCtx();
-    int
-    vramUpdateCtx(void *address, uint64_t dev_id, bool &restart_reqd);
-
     // Memory management helpers
     nixl_status_t
     internalMDHelper(const nixl_blob_t &blob, const std::string &agent, nixlBackendMD *&output);
@@ -300,14 +280,8 @@ class nixlUcxEngine : public nixlBackendEngine {
     std::string workerAddr;
     mutable std::atomic<size_t> sharedWorkerIndex_;
 
-    /* CUDA data*/
-    std::unique_ptr<nixlUcxCudaCtx> cudaCtx; // Context matching specific device
-    bool cuda_addr_wa;
     mutable std::optional<size_t> gpuSignalSize_;
 
-    // Context to use when current context is missing
-    nixlUcxCudaDevicePrimaryCtxPtr m_cudaPrimaryCtx;
-
     /* Notifications */
     notif_list_t notifMainList;
 
@@ -330,9 +304,6 @@ class nixlUcxThreadEngine : public nixlUcxEngine {
     getNotifs(notif_list_t &notif_list) override;
 
 protected:
-    int
-    vramApplyCtx() override;
-
     void
     appendNotif(std::string remote_name, std::string msg) override;
 
@@ -368,9 +339,6 @@ class nixlUcxThreadPoolEngine : public nixlUcxEngine {
     getNotifs(notif_list_t &notif_list) override;
 
 protected:
-    int
-    vramApplyCtx() override;
-
     void
     appendNotif(std::string remote_name, std::string msg) override;
 
diff --git a/src/utils/ucx/ucx_utils.cpp b/src/utils/ucx/ucx_utils.cpp
index d337e7247..5b1019430 100644
--- a/src/utils/ucx/ucx_utils.cpp
+++ b/src/utils/ucx/ucx_utils.cpp
@@ -456,6 +456,9 @@ nixlUcxContext::nixlUcxContext(std::vector<std::string> devs,
     unsigned ucp_version = UCP_VERSION(major_version, minor_version);
     if (ucp_version >= UCP_VERSION(1, 19)) {
         config.modify ("MAX_COMPONENT_MDS", "32");
+    } else {
+        NIXL_WARN << "UCX version is less than 1.19, CUDA support is limited, "
+                  << "including the lack of support for multi-GPU within a single process.";
     }
 
     if (ucp_version >= UCP_VERSION(1, 20)) {